Conference 7.286::digital

    Well, he would say that, wouldn't he? :^)
    
    Laurie.

RE: <<< Note 4189.0 by DECWET::WHITE "Surfin' with the Alien" >>>

>In this weeks Information Week, Andy Grove, CEO of Intel is quoted as saying
>by the decades end, there will be only two chips left....

What's the definition of being "left": available for retail purchases, or in
dominant/high volume production?

Today Alpha is available but neither dominant nor high-volume (by industry
terms). In five years I would think that the situation would be about the same.
						-- Tom

This item is from the 10/16/95 issue of PC WEEK by Jesse Herst.

My simple question:
How can Alpha gain market/mindshare with out a marketing heaveweight 
partner - IBM/Apple/Motorola, HP/Intel?

***************************************************************************

Shape Up or Chip Out: RISC-y Future for Pair

This is the final installment in our series on CPU architectures. So far we've 
discussed the prospects of Intel (great), PowerPC (good), PA-RISC (subsumed 
into Intel's P7), and SPARC (fair). This week we'll conclude with a look at DEC's 
Alpha and Silicon Graphics' MIPS. A sinking chip? Sadly, the Alpha's future is in 
question. The Alpha is the Ferrari of computer chips, but DEC has not been able 
to translate its performance superiority into market share. The Alpha has a 
minuscule 4.2 percent share, according to IDC, compared with 22 percent for the 
PowerPC and 20.5 percent for the PA-RISC. 

Remember, it costs roughly $50 million to design a new chip and another $50 
million or so to retool the plant. When you are selling at the Alpha's tiny 
volumes, the burden per chip is enormous. DEC has a decent high-margin server 
business, but its best hope is Windows NT. If NT becomes the mainstream 
business OS, then DEC may be able to build a volume business for the Alpha. 

    
    	Often it is amazing how the truth (re:-1) requires so few words.
    
    	Maybe our semiconductor and marketing mavens will get the message
    	that in *chips* price+availability=volume;volume+margin usually
        equals profits if you control costs.
    
    
    		the Greyhawk

It seems to me that the time is ripe for Digital to go mainstream on
Alpha marketing.  There is no such thing as 'bad press' only good press.

The fact that Digital and Alpha is supposedly the new leader in small servers
running UNIX ought to give this company the long awaited 'business-reason'
to do some mainstream marketing targeted at the 'I've got 20k tops to spend
on a server' customer.  This customer will buy through channels, but that does
not dismiss our responsibility to generate demand through corporate ad
inititatives.  Heck, let's get the PCBU and the SBU competing against eachother,
(to generate demand through channels, not directly), would be good for the
company, and be quite a novelty fot the trade press to highlight: 

Top Story: "Digital Business Units Engaged in Ad War"  

Alpha has gained momemtum, Digital is scaled back from 120k employees to half
of that.  Let's be nimble and jump on this opportunity.

I'll take a thousand customers buying multia internet servers over two 8400
customers with 600 page RFP's in hand, ANY DAY.

-Stephen

    God, and he promised! :^]
    
    Internal divisions competing against each other is one of the things
    already affecting our bottom line.  Duplication of effort and resources
    in pursuit of the same client dollar produces lovely intramural games
    but is, by inherent nature, not as profitable as competing with our
    competitors.
    
    								Tex

    
    
    well, so surprise right -- what else would one expect an Intel chief
    to say about his competition?
    
    one another matter... either we can lash out behinds all by ourselves
    (we know can can do ably) or we take advantage of what we can do well.
    
    Have you checked out the good work being done (and deals won) in the
    high-end commercial UNIX space, namely us (with the 8400) and Oracle,
    Sybase and Informix. I would encourage you to contact the DB Market
    Development Group for really good information, and to check out
    the database benchmarks that the Benchmark Center conducts.
    
    This is NOT a volume market for chips, but again this is a growing
    market for Digital which has been increasing winnable. 
    
    
    

	Methinks that Andy Grove in 10 years will be joked about the
	same way Ken Olsen is about his "nobody needs a computer for the home"
	statement. To think that there will be only 2 is short-sighted at
	best.

	I'm not pinning at hopes on Alpha Marketing tho. If someone
	doesn't pull their head out of the sand (or someplace else) and
	get with the program, we can kiss Alpha goodbye in a few years.

	We've got the technology, we can build it. Bob P., it's time
	we turn up the marketing heat.

							mike

    Do Intel still sell i960s ? Do they still make money on them ? Does it
    matter that no-one in PC Week's office has heard of them ? Does it
    matter if Andy Grove's not heard of them ?

    Just tell Andy he needs to get a fixed Pentium in the machine he
    used to run the predictions...

From:	US4RMC::"[email protected]" "Do Not Reply. Bounce-backs only" 16-OCT-1995 09:54:07.98
To:	diode::crowell
CC:	
Subj:	1425 POWERPC 620 PROCESSOR IN SERIOUS TROUBLE - DROPPED IN RS/6000s

POWERPC 620 PROCESSOR IN SERIOUS TROUBLE - DROPPED IN RS/6000s
(October 11th 1995) A year after it made first silicon the whole
PowerPC 620 effort is entangled in severe debugging problems. There
are even rumours running around IBM that it will be canned entirely.
Already the chip has disappeared from all IBM's RS/6000 plans.
Speaking at a press briefing in Austin, Texas last week, an
uncomfortable James Thomas, director of RISC microprocessor
development at IBM, said: "While we do have silicon back it has been
a very long, tedious time-consuming debug." Thomas, who is
responsible for IBM's presence at the joint Apple/IBM/Motorola
development lab, refused to give any timescales for the delivery of
the chip, saying only that he is spending a lot of time trying to
collect information on just that question. 

Thomas also refused to comment directly on rumours of the chip's
death. He would only say "We are still actively and heavily in debug"
- not exactly a firm assurance that the chip as currently described
would ever appear. However Motorola director of RISC marketing Phil
Pompa told our sister publication, Unigram.x, that his company is
still fully committed to the chip. Meanwhiel Didier Breton, VP of
Open Systems at Bull is adamant that his company is still "fully
committed to the 620". He admitted, however, that "620-class machines
with SMP will not appear until 1997". Bull was orginally talking
about the end of the year for 620 boxes, but recently put this back
until '96.

So what has gone wrong? It seems that while parts of the processor
are working fine, others are a mess. In particular, Thomas identified
multiprocessing as a continuing bug-bear - the same problem that
bedevilled the PowerPC 604. In general, the Somerset team seems
bogged down by the complexity of the processor, which is designed to
contain 7 million transistors (though many of those are in the chip's
64k cache). Thomas said the chip was "infinitely more complex" to
debug than its predecessors. 

The loss of the 620, should it occur, would be a double blow for the
PowerPC consortium. At the theoretical level it would reflect poorly
on Somerset, which has been painted as one of the world's finest
collection of processor-designing talent. More practically, it would
leave the PowerPC program temporarily without a 64-bit chip, and
without a processor capable of taking on the next generation of
Alpha, MIPs and SPARC chips. But even when it was first announced a
year ago, the PowerPC 620 Spec92 figures looked relatively
lack-lustre compared with these other RISCs. 
(c) PowerPC News - free by mailing [email protected]

% ====== Internet headers and postmarks (see DECWRL::GATEWAY.DOC) ======
% Received: from mail2.digital.com by us4rmc.pko.dec.com (5.65/rmc-22feb94) id AA16374; Mon, 16 Oct 95 09:37:10 -040
% Received: from power.globalnews.com by mail2.digital.com; (5.65 EXP 4/12/95 for V3.2/1.0/WV) id AA25315; Mon, 16 Oct 1995 06:26:51 -070
% Received: (daemon@localhost) by power.globalnews.com (8.6.12/8.6.4) id NAA20537; Mon, 16 Oct 1995 13:24:13 GMT
% Date: Mon, 16 Oct 1995 13:24:13 GMT
% From: [email protected]
% Message-Id: <[email protected]>
% To: diode::crowell
% Reply-To: [email protected] (Do Not Reply. Bounce-backs only)
% Subject: 1425 POWERPC 620 PROCESSOR IN SERIOUS TROUBLE - DROPPED IN RS/6000s

    
    I'm looking for the rest of the notes showing how Intel botched the P6
    design as well....
    
    So they are both in trouble... We are so far ahead in many ways but
    we still are having trouble getting any air time... They don't
    even mention Alpha in the powerPC news files..
    
    Jon
    

Re:
"We're still having trouble getting airtime"

It's pretty simple:
If you agressively market, spending what it takes, like Intel is doing,
your name and concept becomes common knowlege.  News stories tend to reference
common knowlege, even if only for comparison.  Thus, your advertising money
is multiplied.

If you don't spend any advertising money up front (to speak of), there's 
nothing to multiply.

Without our name and image in front of people, we'll never gain mindshare.

Kevin

>>It's pretty simple:

Apparently it isn't.

I'm not even 6 months old at Digital and I'm already tired
of crying about lack of marketing.  A black and white ad on
page 354 of a 800 page PC rag just don't cut it.

This really a new Digital? How 'bout we introduce ourselves then?

Why oh why do we REFUSE?...TIME and TIME again!?!

I know it's not a money issue anymore, not with the latest
salvo of bonuses that got handed out.

We OWE it to ourselves, our company pride, our organizations,
to employees who have not seen a raise in years, to people
who emphatically defend Digital and 'closet-market' whenever
possible, to go mainstream on advertising.  Host a mini-series or 
something, sponser athletic events...put an Alpha in the pit of
an F1 race car team, help sponser a Hydroplane team, build a Digital
Blimp.  I don't now, but for G*d sake, let's shed this no-name computer
company and the 'Alpha what' syndrome once and for all.  

Geez, maybe then I can at least accept how late some of our software actually
ships relative to it's plan released dates.  I would actually LOVE for Digital
to get accused of being all marketing and no substance for once.

Sorry, I'm probably getting to sound like a broken record by now.

-Stephen

    The Digital Blimp
    
    I like it . . .
    
    Bill

  HP would run ads deriding "64-bit bloat"!

                                   Atlant

    
    Wasn't it P.T. Barnum who said he didn't care what people said about
    him in print as long as they spelled his name right?
    
    I doubt if Met Life worries about the connection between Snoopy, a
    blimp and life insurance and if some competitor will say that Met's 
    like a blimp, an empty bag full of gas.
    
    Thousands see the blimp on TV every weekend and they know what (who?)
    Met Life is. I'm not saying Digital needs a blimp but the only thing
    more frustrating to many, many DEC (OOOP...Digital) engineers than
    having the technology and not being able to sell it is having the
    technology and not TRYING to sell it.
    
    If the corporation can afford mega-thousands bonuses to the top
    officers (no value judgement here one way or the other) then it could
    afford to spend a few hundred thou or more on getting our name out.
    
    It could. Will it? History is not encouraging...
    
    
     Gary 

    Digital's Alpha program suffers from all of the aforementioned
    marketing and volume problems, but there's another real problem
    in the UNIX server space: Digital's software. First Ultrix, then
    OSF/1, both packages were weak, non-competitive, and roundly derided
    by the press and the user community. I don't credit HP's success in
    the UNIX market with PA-Risc, I credit HP-UX and the third-party
    support it received.
    
    Digital UNIX may be better, but it's too late, we've burned too many
    customers and our reputation. And VMS is on the downhill slide, so the
    only hopes left to the Alpha program are the "cheap" operating systems,
    like Windows-NT and Linux. And I doubt that we will become the market
    leader over Intel for either of these operating systems.
    
    The bottom line is that our high-end business cannot fund new processor
    development by itself. As mentioned in the previous news article, it
    costs at least $100 million bucks to create a new high-end MPU. How can
    that cost be absorbed if you only sell a few thousand units, even at a
    ridiculous markup??
    
    Geoff

    Take a look at the November 1995 issue of BYTE.  This issue has feature
    articles on CHIPS, and for once a mainstream publication acknowledges
    that Digital and Alpha exist.  --Bob

re.14:

>an Alpha in the pit of an F1 race car team

Digital is an official supplier to the Rothmans Williams Renault team. 
Trackside they use 4 x Celebris 590 PCs.
For CAD they use 24 x Alpha workstations (mostly DEC 3000 model 
400/600); 2x AlphaSrevr 2100; 77.4 gigabytes of StorageWorks RAID.
For suspension testing they use an AlphaStation 200 4/233.

Dave.
(details from Uk edition of Digital Today)

RE: .20

	And the only place one would ever hear about it is in this
	notesfile.

	We need to hire some Microsoft marketing folks. People may
	end up hating us, but at least we'd be well known and 
	successful.

							mike

Byte loves Digital and it's no secret...

I'm glad for Byte, but they always toot our horn for us.

Anybody see the irony in that post about the F1 race car team?

I follow F1 racing very closely, first time I ever heard of this:

Indulge me for a minute:

Camera angle is an overhead view of a bunch of suits riding in go-carts
that are way too small for them on an F1 style track, they look ridiculous.
Announcer silver-toungues something about how our competitors 32 bit platforms
don't cut it.  The Rothmans Williams Renault F1 race car peppered with
Digital logos comes SCREAMING by and literally blows the little pathetic
go-carts off the track.  More silver-tounge about 64 bit platforms and the
end-of-life architectures of our competitors voiced over a scene of these
suits in the garage trying to mount and engine that is bigger that the
go-gart on the chasis, and arguing...

Whole thing ends with the Digital logo zooming into the screen...the letters
'stack up' and bump into eachother from the sheer force of stopping to pose
from the camera...and then zoom off in a delayed effect, the 'l' zooms off
and the rest have to kind of catch up...

And run the damn thing prime time...

    I have some questions for the "SLT readers" or those who digest and
    report to the SLT the gems offered in this conference.
    
    1) What is our corporate advertising budget?
    2) Are we *EVER* going to do any *REAL* prime-time TV advertising?
    
    The concept of the marketing folks putting TV adverts on 'hiatus'
    is unacceptable. I think the ideas expressed in the conference are
    head and shoulders above anything I've seen DEC put on TV so far,
    
    3) Why aren't some of these ideas acknowledged/used? 
    
    -Or is the concept of a humble <mumble>-engineer THAT YOU ARE ALREADY
    PAYING actually having better ideas than the advertising firm that you 
    are paying millions to just stick in your craw.
    
    4) What response are we going to show (this week or next) that replies
       to the new WSJ/INTEL ad that swears INTEL's processors will run
       2-BILLION instructions per second in one <n> years?
    
    These are serious questions, that I as a stockholder first and employee
    seconds would like answers to.
    
    
    .mike.

    From the SBU "Competitive Newsflash":
    
    1. PowerPC Shorts Out (INFORMATIONWEEK, 8/21/95).
    	- new chip fails to spark large sales
    	- Pwr Series 830 & 850 not meeting IBM's sales forecast.
    
    2. IBM is preparing its PowerSeries NT servers for Q1 1996 release.
    	- "Combination of PowerPC and NT is unlikely to succeed".
    	   (Scott Winkler, VP, Gartner Group - PC Week, 8/21/95).
    
    It is not all good news for IBM/Motorola.
    

         <<< HUMANE::DISK$CONFERENCES:[NOTES$LIBRARY]DIGITAL.NOTE;1 >>>
                        -< The Digital way of working >-
================================================================================
Note 4114.0                Digital, Intel & Microsoft                 24 replies
MIMS::SANDERS_J                                      47 lines  11-SEP-1995 11:19
--------------------------------------------------------------------------------
The Atlanta Constitution
Sunday, September 10, 1995
Bill Husted, Staff Writer

"Pentium a tough act to follow"

   Don't junk that Pentium yet.
   Even techno-junkies who find it difficult to live with 
anything less than the latest and greatest are worried about 
Intel's planned successor to the Pentium chip.  The new 
king-of-the-hill chip --- so far called by its working name, P6 
--- is expected to be on the market by the end of the year.

   But early test reports of the prototype chips have been 
disappointing.  Intel has said the new chip will be twice as fast 
as the fastest Pentium.  But the test reports have shown just a 
15 percent improvement ... not nearly enough.

   John Hastings, president of Atlanta-based American Computer 
Exchange, reports in his faxed Computer Trends newsletter that 
many experts feel the "x86" type chips --- the term the industry 
uses to describe 80286, 80386, 80486 and Pentium chips --- may be 
near the end of the line.

   "If the performance of the P6 chip cannot be improved 
dramatically, Intel may be worked to rush its P7 [successor chip 
to the P6] to market sooner than it had intended," Hastings said.

   The P7 will use a different type of internal architecture than 
the x86 line of chips.  It will work more like the PowerPC chip 
used by Apple and IBM.  The PowerPC uses RISC (reduced 
instruction set computing) architecture.  Here's why that 
matters.

   Microsoft's Windows 95 was made to work with the x86 series of 
Intel chips, and the company has said it won't be rewritten for 
RISC chips like the P7.  That would mean, Hastings said, that 
Windows 95 could be a short-lived operating system.

*****************************************************************


Is John Hastings blowing smoke or is he on to something?

What does this mean for Digital?  Intel?  Microsoft?

    read about our P6 box in the Oct 16th issue of PC Week, which hit
    the street today.
    
    It blazes past 21064A-based Alphas with NT for about the same price
    (or even less), or offers near 21164 performance for dramatically
    less money.  It isn't meant for WfWg or Windows95 tho; that's P54C
    territory.  Kratz

More adverstising would be nice...

But, do we really believe that if you "throw money at it, they will come?"

I suggest that something more has to be done  - and soon!

We need to partner with some company to give Alpha the "look" of a collaboration 
like Apple/IBM/Motorola or HP/Intel.

We better move fast, 'cause there are few BIG players who don't already have a 
partner.  Who's left?  Sun, Compaq (although Intel minded for sure) or MIPS 
(probably not willing to swallow their ego).  Are there others?

We always talk of picking the right channel of distribution.  In this case, the 
channel is the right Alpha partner.  We would'nt have to hire anyone and we 
would enjoy the benefits of not looking like a lone ranger with not enough volume 
to make a worldly difference.

Tony

    
    This discussion reminds me of an article that I saved which appeared
    in the June 29, 1994 Digital Review.  The conclusion was that mere
    chips without all the other pieces, are a poor commodity to bet the
    er, "farm" on.  
    
    Enjoy!
    
    
    ----------------------------------------------------------------
    
    DIGITAL REVIEW				June 29, 1994
    



                       BOB 
                 
                     and the 

                    DEC STOCK

                    
          A Parable for Our Time
           


  Note: Any relation to persons, living or dead, or to any company, 
    (living, or slowly dying) is purely coincidental. This is only a 
    fairy tale. Especially since it is Grimm. The only place where 
    fairy tales get confused with  reality is Washington, D.C.
    




One day, not long ago there lived a Wise old man known far and wide
as Uncle Ken. He lived on a farm carved out of a wilderness, where others
later came and thrived, and his products
were sold far and wide.

Eventually, as happens in all things, the time came to hand over the reins
of leadership to the heir apparent of this great empire : Bob.

Ken passed off the corporate scepter, the keys to the Fast Red Wagon, and
the designated parking space. But most important, Uncle Ken handed over
The Cash Cow, The Bread&Butter and the Seed Corn. 

"Of all the trappings of wise corporate stewardship", Uncle Ken told Bob, "the
latter three are  most important. They are the future of our family. It is up
to you to get them to market so that we can profit."
  
So Bob hitched the Cash Cow up to the corporate wagon, loaded up the Seed Corn
and the Bread&Butter and set off down the Road to Market, which as we all
know, is full of twisty turns, all alike, where it is easy to lose your 
purpose. But that is a tale for another day.

As the road grew ever winding Bob began to wonder if he would ever reach
his destination. He encountered many other farmers on the road, mostly going
the other way. He would ask them: "Excuse me, but is this the way to
Profitability ?" They would answer various things like "TQM !", 
"Team Empowerment !", "Vision !", "Managing Valueless Diversity !" and so on. 
And they hurried down the road the way Bob had come.


As the road got steeper, and the Cash Cow started to have difficulty 
pulling the wagon, Bob decided to lighten its load. He stopped the wagon,
causing much confusion, and went to see what the matter was. 
He asked: "Why is the Cash Cow struggling so ? "
From the workers riding in the corporate wagon he heard:
  "Lack of Vision ! Excessive Overhead ! Burdensome Bureaucracy !"
The managers riding in the corporate wagon said various things like:
  "Core Businesses ! Too Many Workers ! Rightsizing !" 
 as they pointed fingers of blame at each other.
So he jettisoned some of the workers and gave them each a little 
seed corn to survive on in the wilderness.

The managers approved of this decision. They awarded themselves to a 
helping or two of the seed corn for their vision and intelligence.

The road got no less steep, but shed of the excess weight, they continued
to inch forward. A passing tradesman named Wall Street advised Bob, 
"If you don't shed some weight, you never get over the hill to Profitability."

So the wagon stopped again, and Bob went off to see what could be done.
"We are not making good progress ! What can be done ?"

Once again the cycle of blame and excuses was repeated. And once more workers
were jettisoned from the corporate wagon and left in the wilderness with a
little seed corn. And the managers talked of eating the Cash Cow in
celebration of their Wisdom and Ability, and ended up arguing over who was
the wiser. One held a moist finger up to test the wind. Another argued that
there was no reality except perception, and that his  perception was superior
to all others. Another sat around and talked of the 'vision of the elephant'.
Meanwhile, Bob, the Cash Cow, and the workers soldiered on.

Eventually he came to a crossroads. A man stood there with pockets full of
gold, and a hat with the brim pulled down low over his face.

He said to Bob: "You'll never make it to Profitability pulling that load."
"I'll trade you your Bread&Butter and the workers attached to it for one of
these magic chips. These chips are made of rarest silicon, and they run very 
fast. If you plant them correctly the will grow into a Stock that will
reach for the sky."

So Bob agreed and the Bread&Butter was unloaded, and some workers with it.

The man at the crossroads looked upon the wagon, with the managers in it 
squabbling over who would manage what was left, and told Bob:
"Your wagon is still too heavy to make it to Profitability. Would you 
trade your Cash Cow for another of these magic chips ? "

Bob agreed.

The managers were aghast ! 

"Whose back will we ride if the Cash Cow is sold?"
"Whose ideas will create another ? "
Terrified, they looked at each other for a vision or and idea they
could claim as their own. But none came.

So a compromise was reached. Some of the managers split up the Seed 
Corn that was left and went their separate ways, except for a few who 
waited around for someone to follow. One was still babbling about a
'view of the elephant'.

The corporate wagon was abandoned along the side of the road.

All that was left of the once mighty family was Bob, a few managers, and a
couple of magic chips.

Bob and the few stragglers made it over the hill to Profitability and
found a field of rich loam ripe for the planting.

Bob dug a hole in the ground, and planted one of the magic chips.
The other, he handed to one of the managers who had been tagging along.
The manager dug a hole, planted the other chip and looked up with a smile:
"I can do that too, now that I see how it is done."

And they all sat around and waited for the DEC Stock to grow.

And they waited. 
And one manager talked of 'perception'.

And they waited. 
And one manager talked of shades of gray.

And they waited.
Another talked of 'views of the elephant'.

The last I heard, from a friendly entrepreneur who passed them by on 
the way to Profitability, they are waiting still, as others develop marketing
channels to deliver their products, and team up with VAR's, or offer 
bundled solutions.

You see, a chip is just so much silicon, unless it is backed up with 
Bread&Butter, which some call software, a corporate wagon, which some
call systems integration, and a Cash Cow, which some call
service, and support.

   The End
    
    * The author, when not amusing children with his fairy tales, lives in 
    a castle headquartered in Maynard, Mass.

PC Week October 16, 1995 Volume 12, Number 41

OBJECTS: Pact with Software AG signals dissatisfaction with DEC

Microsoft Corp. has contracted with Software AG to port OLE to a variety 
of UNIX platforms and to IBM's OS/400 and MVS operating systems over the
next three years.

Software AG is slated to port COM (Common Object Model) and Network OLE
(Oblect Linking and Embedding) to all major UNIX platforms and to OS/400
and MVS.  To accomplish this, the Rseton, VA. company will implement OLE
in its Entire client/server middleware product line and in its Natural
fourth-generation programming language.  The company will also offer its
customers OLE integration and support services, officials said.

The Software AG relationship to a large extent overlaps the current 
Microsoft-Digital Equipment Corp. deal, announced at the end of 1993, to
port COM to a host of UNIX, microcomputer, and mainframe operating systems.

Some industry watchers said Microsoft is dissatisfied with DEC's inability
to deliver COM on these platforms and that Microsoft has been gradually
weaning itself from DEC as its primary OLE porting partner.

DEC officials acknowledged that the company has been slow to make cross-
platform COM available to customers.  "We're behind where we thought we
would be two years ago," said Steve Baron, senior software engineering
manager for Object Broker, DEC's Common Object Request Broker Architecture
middleware product, in Nashua N.H.  "We haven't yet shipped anything on
the server side, although we have some very early pilot kits avaialble".

By mid-1996, Software AG is slated to provide OLE automation for Windows 
clients by integrating MVS, Empire and an OLE interface, officials said.
The resulting technology will let clients "wrap" legacy applications with
front-end object technology, making legacy mainframe data or transaction
services available in the form of server-based object components.

By the end of next year, Software AG plans to ship an OLE- and OLE Custom
Control-Enabled version of Natural, with a Network OLE-enabled version to
folow by 1997 and 1998, the company will offer  native Network OLE on UNIX,
OS/400 and MVS.
    

    	Regarding advertising, we have to keep our priorities straight. I
    mean, how are we going to keep giving hundreds of thousands of dollars
    in salary increases, bonuses, and stock options to upper management if
    we keep throwing away money on silly things like advertising? Get real.
    If we keep patting ourselves on the back long enough, everyone will
    eventually be beating down our doors for our products. Don't worry.
    	A number of years ago, I remember pushing really hard to meet a
    major milestone...working extra hours...weekends...and management
    really came through. I was beaming when I received that $75 cash bonus!
    Take the family to dinner, they said! Wow, it was quite a thrill. I get
    chills whenever I think about it. It's motivators like that which keep
    me fired up and giving 110% all the time. Some of you may be annoyed
    because you've never received a $75 bonus and haven't had a salary
    increase for several years, and maybe it doesn't seem fair when you see
    top brass getting 20% salary increases and hundreds of thousands of
    dollars in cash and stock bonuses. But don't worry. If you keep giving
    it all, you'll eventually get what's coming to you. Trust me.

    re. 30
    
    What does this have to do with Andrew Grove's comments about Alpha?

    RE:.-1
    
    Could it be something to the effect of;
    
     "Others say things in public about our stuff, while we don't?"
	
    .mike.
    
    (Still waiting for the 3-marketeers to do cool_stuff on TV)

Is that we are getting a reputation for being extemely slow (mostly with
software deliverables), no nimbleness whatsoever, and the we are just plain
deficient when it comes to marketing ourselves.

These things will hurt us big time.

We need to address these two areas if we truly want to recover
and begin growing again.

-Stephen

    In the June edition of UNISPHERE, the magazine for Unisys users, there
    was an article on the new Unisys Open Parallel Unisys Server (OPUS), a
    system that uses Intel Pentium chips and Intel's MESH interconnect
    technology.
    
    The article included the text of the press conference that was held at
    the OPUS announcement in New York.  Andrew Grove was one of the
    participants.  UNISPHERE magazine asked all the questions.  The staff
    of UNISPHERE, because of my groups efforts, is very well versed in
    Alpha and Oracle VLM.
    
    Grove was asked the following question:
    
    "How long before OPUS will take full advantage of 64-bit addressing
    capabilities and what sort of price/performance ratio can OPUS realize
    vs. a 64-bit system running ORACLE?"
    
    His answer:
    
    "Performance can be achieved by scaling the ..... frequency of
    architecture running a single processor, but the really dramatic
    element of today's announcement ..... is a simultaneous working of a
    multiplicity of processors, up to 128 of them.  When you do a little bit
    of multiplication and take 128 processors, each of them operating at 250
    to 300 MIPS, you get easily 30,000 MIPS performance."
    
    
    Now is that avoiding the question or what?  What a bunch of B___S____!
    
    In a recent Oracle benchmark held for a customer, Tandem beat OPUS, and
    the AlphaServer 8400 was more than twice as fast as Tandem.  The OPUS
    system was using 32 processors.
    
    Andrew Grove avoided that question because he DOES NOT HAVE A GOOD
    ANSWER!
    

Of course not. The Intel 80xxx architecture is at a dead end. I reckon 
it has 5 years life at the MOST. If it wasn't for DOS/Windows, who 
would buy it !


Alan

    Oh, I wouldn't count x86 as dependent on DOS/Windows... P6 is
    amazing running NT (and faster than 21164 on some native NT
    benchmarks).  P6 clobbers 21064A and P6 systems will cost the
    same or less.  P6 should effectively keep Alpha from making
    any serious inroads to the desktop NT market over the next
    few years.  .02 Kratz
    

    re. 36
    
    Can you provide the numbers and the type of benchmarks run?
    

    .35
    
    Our arrogance shows through again. We have always had (in most cases)
    superior technology. Does Digital sell the most...NOT..  marketing
    has a lot to do with success of a company.
    
    .35 must be an engineer.
    

>              ... P6 should effectively keep Alpha from making
>   any serious inroads to the desktop NT market over the next
>   few years.  .02 Kratz

    P6 keeping Alpha out of the desktop market???  You've got to be
    kidding.  Our lack of marketing and is *far* more effective at that
    than Intel and P6 could ever hope to be ... ;-(

    Bob

From recent articles in BYTE, et al, it would appear that the P6 has the
same flaws as Alpha (i.e. an architecture in place before the byte-orientation
necessity placed by the success of Windows). They claim that P6 performs
below existing Pentium systems in many applications.

I think the battle over which chip will dominate the Desktop NT market will
be decided on how well they handle existing software (written for Intel 86
architecture). If the press reports are correct, Intel appears to have
opened up a window of opportunity for us here. PErhaps this is what is
behind Andy Grove's comments? Just trying to talk his way out of the error.

Mark

Mark:

> From recent articles in BYTE, et al, it would appear that the P6 has the
> same flaws as Alpha (i.e. an architecture in place before the byte-orientation
> necessity placed by the success of Windows). They claim that P6 performs
> below existing Pentium systems in many applications.
> 
> I think the battle over which chip will dominate the Desktop NT market will
> be decided on how well they handle existing software (written for Intel 86
> architecture). If the press reports are correct, Intel appears to have
> opened up a window of opportunity for us here. PErhaps this is what is
> behind Andy Grove's comments? Just trying to talk his way out of the error.

  There's a crucial difference, though, between the P6 situation
  and the Alpha situation. P6 still includes the byte- and word-
  oriented instructions of the x86 architecture; it's just that
  the implementors chose to de-emphasize byte and word performance
  for better performance with the larger operands.

  SHOULD INTEL DESIRE, they could easily re-spin the chip with
  a better performance balance between the instructions, and people
  buying the re-spun chip would get an immediate performance benefit.

  Our chip *HASN'T GOT* any byte- or word-oriented instructions so
  no Alpha software uses such instructions. Even if we introduced
  a chip with newly-architected byte and word instructions, there
  would be very little performance gain until software caught up
  with the new chip, assuming it ever did.

  Our omission of byte- and word-oriented instructions from the
  Alpha architecture was a serious mistake which can not be easily
  rectified at this time.
                                   Atlant

	ugggg...first off, the 21164A will have byte/short instructions...
	as to whether that, or their omission from earlier Alpha
	generations, really matters is a HIGHLY religious issue,
	and certainly one that does not boil down to pat statements
	such as -.1...the perception problems around Alpha's performance
	have more generally been due to software issues (compilers,
	OS optimizations, the delay of SoftWindows V2, etc) than
	hardware ones...

	IMHO of course...

	_kelley

    RE: .42 & .41

    IMHO, the performance tradeoff chosen (NO BYTE/WORD) when the Alpha
    architecture was being debated missed the fact that the rest of the
    world still revolved around I/O which was (and still is in many cases)
    byte and word oriented.  Sure, we've got a killer chip which can crank
    out performance in the server market.  That's where our successes have 
    been.  However, if you want to get I/O into an Alpha system and the
    world you're connecting to isn't a new design, then you'll experience
    what .41 is talking about!

    ... mailboxes ... dense space ... sparse space ...  blank space ...

    Kevin

  Kevin, thanks. As you know, I/O performance is a big part of my
  bread-and-butter in Embedded and Realtime Engineering, and Alpha's
  lack of byte- and word-oriented instructions is a killer on legacy
  busses like the VMEbus and ISA and pseudo-legacy (outgrowth) busses
  like the PCI, where many devices still have dependencies on being
  able to issue reads and writes to specific bytes. LCA's kluge of
  "sparse space" seems like a bad joke to many people.

  I also was also was thinking about legacy codes, many written in
  machine language or interpretively emulated.  (In other words, x86
  and MC68K shrink wrapped codes, among others)

  Alpha's lack of byte- and word-oriented instructions is deadly
  for the perfomrance of these codes. See the ONTIME::MULTIA con-
  ference if you have any doubts.

  If you don't already have this conference in your notebook and
  would like to add it, press <KP7> or <Select> or type "SELECT"
  and the conference will be added to your notebook.

                                   Atlant

Kelley:

> ugggg...first off, the 21164A will have byte/short instructions...

  I don't think that's "public information". You might want
  to check, and possibly, revise your note.

                                   Atlant

Kelley:

> the delay of SoftWindows V2, etc)

  Gee, SoftWindows V2 (full '486 emulation) has been running for
  awhile, in the customers hands, on PowerPC. Of course, they
  *HAVE* byte- and word-oriented instructions. That wouldn't
  have anything to do with it, would it?

                                   Atlant

re:    <<< Note 4189.46 by ATLANT::SCHMIDT "See http://atlant2.zko.dec.com/" >>>
              -< No quarter given on *THIS* issue (bytes/words)! >-

	see, I tolda ya it was a religious issue ;-) I never said that
	lack of byte/word didn't create problems in some sectors...
	I only said that other lackings, mostly in the software realm,
	are IMHO the more visible issues keeping Alpha out of the
	mainstream...

	it is my understanding that the lack of byte/word (as well
	as divide, square root, etc.) in the first generations was
	a well considered trade-off...perhaps those making said decisions
	would do differently now with hindsight, perhaps not...

	re SoftWindows V2: I'm not in a position to comment on Insignia's
	porting priorities but MACos versions seem to generally precede
	all others...SoftWindows V2 on Digital UNIX is imminent (as is,
	I am led to believe, new 486-based underpinnings for Alpha NT)...

	re 21164A info...I got that from another "public" conference,
	RICKS::DECHIPS....since this info was presented at the
	1995 Microprocessor Forum, I assume its already in the public
	domain.....

	_kelley

_kelley:

> re 21164A info...I got that from another "public" conference,
> RICKS::DECHIPS....since this info was presented at the
> 1995 Microprocessor Forum, I assume its already in the public
> domain.....

  DEChips 439.11 says the PowerPoint presentation (from which you
  read in 439.10) is marked "Digital Confidential". I'm not aware
  we're generally discussing the ECO yet.

                                   Atlant

As the text you specifically pointed out mentioned, this information was
announced at the public Microprocessor Forum.  Specifically, the talk
mentioned the 6 instructions being added to the 21164A and all future
Alphas: {load/store/sign-extend}{byte/word}.

    Kelly,

    The lack of byte/word in Alpha isn't a religious issue, it's based on
    fact.  I don't disagree with you that it was a well considered trade-off.
    Rather, I believe that it reflects the arrogance (of the time) that
    Digital would conquer the world with blazingly fast hardware, in total
    disregard for what customers would need to do to migrate from existing
    I/O systems.

    With few exceptions (storage and networking come to mind) Digital made
    the decision to treat all I/O (Digital produced or third party) as a
    third party problem, and provided no simple means of getting the 
    benefits of the Alpha architecture while integrating with the real 
    world surroundings of our customers.  If you've ever had to develop or
    port I/O related code to an Alpha system you would know exactly why
    software performance is lacking in that area.

    Kevin

    re .40
    A P6 150 runs 16bit WfWg/DOS at about the pace of a Pentium 133,
    or still quite a bit faster than Alpha.  In other words, P6 gives
    @50% faster than EV45 native NT performance, @50-100% faster than
    Ev45 Win16 performance, and it costs about the same as the NT-only
    version of Ev45.
      
    P6 can also run multiple operating systems at its price; the Ev45
    variant is NT-only.
    Kratz

    So, what am I to do? Buy INTEL stock and send out resumes??
    
    

    re: .52
    
>   So, what am I to do? Buy INTEL stock and send out resumes??
    
    So far, INTEL stock has proven to be a sure bet. As far as the sending
    out resumes part, that's worked well for some, less well for others ...
    
    The sad truth is that technical superiority isn't a magical cure. It
    astounds me after all these years and changes at DEC that we still have
    the "engineer it and they will come" mindset. If it's an incomplete
    solution (no software), and you can't build it in volume (witness our
    incredible backlogs), and you can't offer it at a competitive price,
    then you will not truly stop the downward spiral.
    
    I recently bought a high-end Pentium and Windows-95. I never considered
    buying an Alpha or Windows-NT, because of two things: the software I
    own, and the software I intend to aquire. With thousands of dollars
    invested in software that only runs reliably on x86 architectures, the
    cost and performance of the CPU was a secondary concern. And because my
    future software purchases will be made off-the-shelf or mailorder from
    mainstream vendors, I doubt that I will ever run across any new stuff
    that runs on or requires Alpha-style CPU's. Hence, I join the vast
    majority of computer purchasers out there who look to Intel and MS for
    the forseeable future. Not by choice, just by necessity.
    
    Geoff

>>    So, what am I to do? Buy INTEL stock and send out resumes??

Buying Intel stock may not be a bad idea, but hold off on the resume!

Alpha was designed to be the fastest chip in the world, it seems to have 
taken that position and be holding on to it. In the compute-intensive and
Server applications that shows. 

We are selling more and more Intel based stuff, se we are finally leveraging
off Intel's hold on the desktop. Nwtworking seems to have recovered it's market
share. We do seem to be well into a successful
transition to a market led Company. I think things are looking good for
Digital right now.


M 

    
    Intel will most likely remain the desktop appliance of choice for most
    people, whether they are in the office or at home.  It's fast,
    inexpensive and there is *LOTS* of software that will run on it.
    
    Now, moving up the food chain, someone has to build servers to store
    information.  And that someone is us.  We build the best, fastest most
    inexpensive 64-bit servers in the galaxy.  Even Bill Gates will tell
    you that.  Intel designed their very first chip on a VAX and they made
    it the same architecture as the VAX, called little Endian.  Our VAX's
    are little Endian, the Alpha is little Endian and this is very very
    good news for us.  Why?  Because of the following.
    
    If Im going to do true client-server computing I want a tightly
    integrated architecture between my client (Intel box) and my server
    (someone's box).  So, the tightest integration you can have is if both
    are the same architecture, namely little Endian which is most
    significant digit on the left, just like your calculator or how you
    write numbers in your check book.  All the other manufacturer's (IBM,
    SUN, SGI, HP, Tandem, AT&T, etc..) make their chips Big Endian.  That
    is the big numbers to the right.  So, for them to handle client-server
    computing the servers have to constantly flip the info back and forth
    between the server (big digit on the right) and the client (big digit
    on the left).  This consumes horsepower on the CPU's, horsepower which
    normally would go to working on problems *and* applications.  
    
    Now, you say, "Hey, IBM ain't no dummy they are not going to loose the
    C/S battle over some misnomer as to where the number 1 is stored on a
    computer !" and you are right.  Which is why all their new Power PC
    chips are Bi-Endian.  You switch them back and forth in the silicon. 
    But here's the rub:  Their version of UNIX, AIX, is Big Endian so they
    have to go back and R&D it to make it work with a Bi-Endian chip.  Big
    Problemo and Big Money.  And they don't quite know what to do with NT
    yet do they?  Nor does HP, their in the same boat UX ain't 64-bit UNIX
    folks, that's why they can't call it UNIX.  It's not SPEC 1170 yet
    (lot's of R&D and $$'s here again too) and well they've got their hands
    full.  
    
    So what have we got?  
    
    We've got the best, fastest UNIX in Industry,SEPC 1170 compliant and all, 
    that's why we call it Digital UNIX.
    
    We've got the best, fastest and cheapest 64-bit NT server and
    workstations in the Universe.
    
    We've got seamless integration between the desktop and the server, no
    flipping the data back and forth which chews up lot's of computer
    cycles.
    
    We've got VMS Affinity with NT.  Clustering in NT and executable code
    built on a Alpha and running on an NT server.  Gate's loves this
    because we have millions of users running VMS which he would like
    to sell his stuff to.  
    
    We've got Services to maintain and take care of these environments
    better than anyone else.  Proof point:  Even Compaq and Microsoft came
    to that conclusion when they named us their PC Service provider and
    Worldwide Network Support Vendor respectively.
    
    And we've got really good engineering, partners, sales people, support
    people and administrative people to keep moving this stuff into the
    right peoples hands.  It's darn near good as penicillan and we 'ought
    to have tables set up in the grocery stores handing out free samples
    for everyone to try. 
    
    So this is good news for us.  We should MARKET the living hell of it
    and tell everyone we know the story.  Let's get going...the battle is
    being waged as we speak.
    
    

Little-endian is such a tiny part of the "VAX architecture" - please be careful
with your hyperbole here.

			Steve

    
    	That is what *marketing* is, Steve - hyperbole; artfully packaged,
    promoted, and priced.
    
    	Want to check it out?
    
    	Try Microsoft, SUN, COMPAQ, and H-P. Their marketing is *always*
    better than their products. If you want to play with the big dogs,
    you better get your A** off the porch...
    
    
    		the Greyhawk

RE: .57

	However, Steve is right. If we were to jump up and scream from
	the rafters that Alpha is great because it's Little Endian you
	would experience the following:

	People in the know would say "Big deal".
	People not in the know would say "Huh??"

	We're best to stick with things like "We're faster than anything
	else at solving your problems" and other types of comments. Banking 
	on Little Endian compatibility would not be prudent.

							mike

Besides, Alpha (the chip) is "bi-endian" (with a small bias towards
little-endian).  All of Digital's Alpha systems are little-endian.

				Steve

    it's just another column on the comparison chart, right next to 
    "64 bits?"
    
    Mark

Nit:
> Intel designed their very first chip on a VAX 

Actually, it was on a PDP-8, in 1968.  Their first and second
chips were a 16x4-bit bipolar RAM (think register file) and
a 256-bit MOS SRAM.  Their first big success was the 1103, a
1024-bit MOS DRAM (3-transistor cells, I think).

Their first microprocessor was the 4004 chipset, a stranger beast
than you are likely to imagine.  Certainly not byte-addressed.

I don't think the 8008 microprocessor had any 16-bit operations,
so the hardware didn't impose any endianness.

I think the 8080 did have 16-bit operations, so it could have
been Intel's first little-endian microprocessor.  If not, the
8086 was the first.  In either case, they more likely inherited
this attribute from the very popular PDP-11 (remember that?) made
by a company called DEC, since the design of the 8086 would have
started just about the time VAX was announced.

  The 8080 had 16-bit operations in at least two senses:

    o Several of the registers (B/C, D/E, H/L) could be treated as
      individual bytes or used with their buddy as 16 bit words.
      Instructions existed which allowed 16-bit loads and stores
      from memory into/from these paired registers.

    o The JMP and CALL instructions could take an immediate
      address from the next two memory bytes.

  Both of these operations, depending upon 16-bit values in byte-
  addressed memory, imply "endianess".

 -=-=-=-=-=-

  By the same argument, the 8008 was actually litle-endian as well.
  While it didn't have an 16-bit load-n-store capabilities, it
  did use a 14 bit address for JMP and CALL, and that 14-bit ad-
  dress was stored in the instruction stream in little-endian
  format.
                                   Atlant

Oh come on. Look at .55 in more depth than one point. I think all 
those messages are GREAT. Now, lets get them made public in 1000' high 
letters !

Alan

re .55
the details of little endian and big endian seem wrong in .55.
the big endian machines have the bits in the same order we write them
on paper.  our little endian systems are the ones that are "backwards".
the difference between the systems is usually not visible to users.
little endian systems have some advantages, including being the way
van Neumann said to do it, but intuitiveness is not one of them.
converting the endianness of a number is not a big problem.
deciding if you have to change it or not is harder.

RE: .64

Big-endian architectures also have some advantages.  From a RISC 
architecture standpoint, one big advantage of a big-endian 
architecture is that both character strings and arithmetic items have 
their bytes in the same order:  you can use quadword-wide compare 
instructions to see if one ASCII string collates before another, 
whereas on a little-endian machine you must compare byte-by-byte (or 
reverse the bytes before comparing).

--PSW

    
    >Digital UNIX may be better, but it's too late, we've burned too many
    >customers and our reputation. 
    
    That's why revenues are up 20% and growth is at 60%over last year 
    with our UNIX product.  Digital Unix hasn't burned anyone, Digital
    as a company has had a few false starts with this market and now 
    does almost 1 billion dollars a year of Unix Business.
    
    >And VMS is on the downhill slide, 
    
    What planet are you reading the returns from?  OpenVMS is moving again
    and has increased revenues over the last 12 months (quarter to quarter)
    
    We are selling a boat load of OpenVMS (3-4 billion last year) and with
    service revenues driving the OpenVMS business to about 8 Billion of 
    profitable, good, solid business with a product that our customers
    still want and potential customers would do well to evaluate for 
    high availablity, production quality compute environments.
    
    >so the only hopes left to the Alpha program are the "cheap" operating 
    >systems, like Windows-NT and Linux. And I doubt that we will become the 
    >market leader over Intel for either of these operating systems.
    
    And we should sell Alpha into the commodity markets so that we can be
    considered and recognized as a full service computer vendor.  This 
    marks us differently than a "Just A Software House" in our customer's
    eyes.  We can integrate the entire enterprise with off the shelf
    products from Digital, Microsoft, Unix, and MVS... Few if any of our 
    competitors can deliver fully on that claim...

    
    
    >Geoff
    
    John W

    re: .66
    
    >That's why revenues are up 20% and growth is at 60%over last year 
    >with our UNIX product.  Digital Unix hasn't burned anyone, Digital
    >as a company has had a few false starts with this market and now 
    >does almost 1 billion dollars a year of Unix Business.
    
    While I won't dispute your numbers, I will question whether our market
    share has grown, especially at the expense of the dominant UNIX vendors.
    Revenue growth is OK too, but are we just growing at the same rate as
    the industry as a whole? In my little corner of the world, I have yet
    to see us take a major customer away from HP, while HP stripped us of
    whole sections of business when Digital was pitching Ultrix and OSF/1.
    
    >What planet are you reading the returns from?  OpenVMS is moving again
    >and has increased revenues over the last 12 months (quarter to quarter)
    
    Same question here too: Does someone think that OpenVMS is gaining
    market share, or that major software vendors are releasing their star
    products first on VMS? Are we selling new customers on VMS, or are we
    living off the installed base and just raising prices to keep money
    flowing for a little while longer?
    
    >And we should sell Alpha into the commodity markets so that we can be
    >considered and recognized as a full service computer vendor.  This 
    
    The point I was trying to make is that Alpha *must* (not should) be
    successful in the commodity markets, or we will not be able to afford
    the cost of designing and building new microprocessors. Intel can just
    plow us under with P6, P7, and so on, and we will bleed red ink trying
    to keep up in the R&D effort. It has nothing to do with being perceived
    as a "full service" vendor. It has everything to do with keeping up in
    the microprocessor arms race without losing our shirts.
    
    Geoff

    Well, at least I now know what the phrase "Little-Endian" means when
    used in this conference. Hitherto, I was relating it to its satirical
    use in "Gulliver's Travels" where it was used to ridicule what we would
    now call a "religious issue", as in, "my editor's better than your
    editor".
    
    Laurie$enlightened.

>    Well, at least I now know what the phrase "Little-Endian" means when
>    used in this conference. Hitherto, I was relating it to its satirical
>    use in "Gulliver's Travels" where it was used to ridicule what we would
>    now call a "religious issue", as in, "my editor's better than your
>    editor".

That is the etymology of the usage.
It's interesting to note that a lot of people do up-case it as Little 
(or Big) Endian, probably (but perhaps not knowingly) in tribute to Swift.
It IS, of course, a "religious issue" with computer architects.
(Swift's reference was to the dispute as to whether one should open 
a soft-boiled egg at the little end or the big end.)

- tom]

    Re .69
    
    >> (Swift's reference was to the dispute as to whether one should open 
    >> a soft-boiled egg at the little end or the big end.)
    
    Why, the little end, of course.
    
    
    

Swift's solution was to open it in the middle...

			Steve

   Re .71:

        Hmm, Middle Endian ?   The odd bits go up, the even bits
    go down ?

      7 5 3 1 0 2 4 6

      0 0 0 0 0 0 0 1   (Converted from little endian)
   
    

>   Re .71:

>        Hmm, Middle Endian ?   The odd bits go up, the even bits
>    go down ?

How about: integers start at the little end and floating starts
in the middle... ;-)  "Scrambled" if you prefer... :-)

Laurie:

  I just happen to have this handy from a document I wrote a while back:

                                   Atlant

 -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

From Gulliver's Travels by Jonathan Swift

  "...Besides, our Histories of six Thousand Moons make no mention of any
  other Regions, than the two great Empires of Lilliput and Blefuscu.
  Which two mighty Powers have, as I was going to tell you, been engaged
  in a most obstinate War for six and thirty Moons past. It began upon the
  following Occasion. It is allowed on all Hands, that the Primitive Way
  of breaking Eggs before we eat them, was upon the larger End: But his
  present Majesty's Grand-father, while he was a Boy, going to eat an Egg,
  and breaking it according to the ancient Practice, happened to cut one
  of his Fingers. Whereupon the Emperor his Father, published an Edict,
  commanding all his Subjects, upon great Penalties, to break the smaller
  End of their Eggs. The People so highly resented this Law, that our
  Histories tell us, there have been six Rebellions raised on that
  Account; wherein one Emperor lost his Life and another his Crown. These
  civil Commotions were constantly fomented by the Monarchs of Blefuscu;
  and when they were quelled, the Exiles always fled for Refuge to that
  Empire. It is computed, that eleven Thousand Persons have, at several
  Times, suffered Death, rather than submit to break their Eggs at the
  smaller End. Many hundred large Volumes have been published upon this
  Controversy: But the Books of the Big-Endians have been long forbidden,
  and the whole Party rendered incapable by Law of holding Employments.
  Now the Big-Endian Exiles have found so much Credit in the Emperor of
  Blefuscu's Court; and so much private Assistance and Encouragement from
  their Party here at home, that a bloody War hath been carried on between
  the two Empires for six and thirty Moons with various Success; during
  which Time we have lost Forty Capital Ships, and a much greater Number
  of smaller Vessels, together with thirty thousand of our best Seamen and
  Soldiers; and the Damage received by the Enemy is reckoned to be
  somewhat greater than ours. However, they have now equipped a numerous
  Fleet, and are just preparing to make a Descent upon us: And his
  Imperial Majesty, placing great Confidence in your Valour and Strength,
  hath commanded me to lay this Account of his affairs before you."

    Thanks Atlant! It just goes to show, there's nothing new under the
    sun...
    
    Laurie.

One could consider the VAX floating types as "middle-endian".

				Steve

I saw the following in the Educom email news. Is it true? Does Digital have
a response to this?

INTEL IS "TERRIFYING" COMPETITOR
The publisher of Microprocessor Report describes Intel's ruthless marketing
strategy:  "I think Intel's business model, put most simply, is that you
build the fabs and then you price the chips to sell out the capacity in the
fabs.  That business model virtually requires that they maintain their
market share, and really the variable is what price they have to sell the
chips for.  I think that's a reasonably terrifying model for competitors to
have to face."  (Investor's Business Daily 23 Oct 95 A8)

And for those who are interested, more information on Educom:

*****************************************************************
Edupage, 22 Oct 95.  Edupage, a summary of news items on information
technology, is provided three times each week as a service by Educom,
a Washington, D.C.-based consortium of leading colleges and
universities seeking to transform education through the use of 
information technology.
*****************************************************************
***************************************************************

EDUPAGE is what you've just finished reading.  To subscribe to Edupage:
send a message to: [email protected] and in the body of the message
type: subscribe edupage Greg Maddux (assuming that your name is Greg
Maddux; if it's not, substitute your own name).  ...  To cancel, send a
message to: [email protected] and in the body of the message type:
unsubscribe edupage...  Subscription problems:  [email protected].

[Posted by WWW Notes gateway]

    Re: .78  "terrifying" competitor
    
    Wooo Wwooo It's Halloween?  Scare them competitors :-)
    

    2. Is IBM's 64bit, PowerPC 620 in trouble?
       --------------------------------------
       Contributors: Bob Welzel @KYO, DTN 323-4380
		     Bob Hendrickson @MRO, DTN 297-3860

    May be. It seems that IBM had long looked to the PowerPC 620 architecture 
    to be the first PowerPC node for it's SP2s systems, but has had to 
    retrench back to it's current multi-chip Power architecture because of the 
    620's poor performance. Specifically, the 620 is unable to out-perform 
    even the recent 77Mhz Power2's 307.9 SPECfp92 floating point performance
    "...and there is a growing cloud over the first 64-bit iteration of the
    architecture, the PowerPC 620."
    	
    This isn't Digital's opinion. That is the measured analysis of London's
    "Computergram", the daily newspaper for data processing, communications 
    and microelectronics professionals and investors in the 9/15/1995 Issue, 
    Number 2750.

    What does all this mean?  Several things;--first a multi-chip 
    design COSTS MORE to build than a single chip, and that means that 
    more often, Alpha will be a better price/performer. Also, this latest 
    news only adds to the ongoing questions about the availability of the 
    PowerPC chip family. 

    Bottom line: Alpha has been here, performing for years, and is only 
    getting better.  We can prove it.  Best performance and best
    price/performance - period.
    	

> Also, this latest news only adds to the ongoing questions
> about the availability of the PowerPC chip family. 

  Yeah, after all, they've only shipped something in excess of
  two or three million units, as compared to Alpha, which has
  shipped, umm, err, nevermind.

                                   Atlant

>> Also, this latest news only adds to the ongoing questions
>> about the availability of the PowerPC chip family. 
>
>  Yeah, after all, they've only shipped something in excess of
>  two or three million units, as compared to Alpha, which has
>  shipped, umm, err, nevermind.
    
    Ah, the joys of 'spin' 8-)
    
    C'mon, don't you see this is just another opportunity for us to
    squander? What difference do the details make?
    
    ...tom

re: Note 4189.0 by DECWET::WHITE 
> In this weeks Information Week, Andy Grove, CEO of Intel is quoted as saying
> by the decades end, there will be only two chips left:
>
> Intel/HP's P7 and P8 and IBM/Motorola's PowerPC.

Interesting.  I was thinking the same exact thing about three weeks ago...

RICKS::DECHIPS Note 288.30  by SUBPAC::MAGGARD  -< Andy Grove's Grand Plan >-

Think about it from Intel's perspective for a minute.

Intel envies the margins that the workstation vendors get on their systems.
Intel wants the same, but they first have to corner the market on PCs and
stomp out all the competition.

They will do so by using their CPU margins as a PC system margin buster.
They're the only ones who can do it.  Every other system vendor has to buy
first from Intel before they can sell to the market.  I think Bill Johnson (DS
Marketing VP) coined the term "unfair competitive advantage."

If Intel drops their PC system prices to a level where only Intel makes a
profit (i.e. CPU profit == system profit), then everyone else dies, and dies
hard.  Then after the barrier to entry is as high for PCs as it is for CPUs,
Intel will just raise PC system prices to a level where they can make the same
margins as us Workstation folks.

Knowing this, it makes me wonder why AMD is not in the PC business.  I guess
they can always buy out Gateway 2000 when Intel kills it.  Whoops, I forgot
about the Intel/AMD FAB-volume agreement.  Too bad for AMD.  They missed that
one, didn't they.

I wonder what the FTC is planning to do about it... 

...

In 20 years, you'll see Intel and MS owning the entire computing market, and
most of us will be working for one of them.  Intel will leverage their PC
volume and high CPU margins to develop high performance workstations using
some next-generation-chip-architecture.  The economies of scale they reap from
their "killer" PC business will squash out all workstation vendors.  Why?  NT.

The box is irrelevant, it will be assimilated.

Alpha (and Digital) will die if 1) we can't keep our significant performance
lead, and 2) if we can't generate a LOT of volume in the next 5 years.


- jeff

    Sounds like the BORG (aka Intel) are approaching faster than expected
    to assimilate us all!
    
    ;^)

After posting -.83 in the DECHIPS notesfile, I received the following response
in e-mail (responses are paraphrased):

> >If Intel drops their PC system prices to a level where only Intel makes a
> >profit (i.e. CPU profit == system profit), then everyone else dies, and dies
> >hard.  Then after the barrier to entry is as high for PCs as it is for CPUs,
> >Intel will just raise PC system prices to a level where they can make the same
> >margins as us Workstation folks.
> 
> But if Intel raises their system prices to contain workstation-level
> margins, then other vendors will get BACK into the market.`, because they
> could once again make a profit at the market price for PC's. 

Perhaps.  But to gain significant market share to be a profitable business,
one has to offer  drop-in replacement that offers the customer a BETTER
overall solution than what they currently have.  And we (Digital) have learned
that there's more to just offering a technically superior product at a lower
price.  People won't buy a computer just because it's faster or cheaper.  It
has to do so with enough margin of benefit that any FUD can be overlooked --
and fear is the saleperson's ace in the hole in the high-stakes computing
business.  

As long as Intel keeps a watchful eye on any possible competition, it can
control it's prices to sustain a significant barrier to entry in the market.
Who are you going to buy your CPU chips from if it isn't Intel, who are you
going to buy your motherboard and chipsets from if it isn't Intel?  Just look
at the PC market today and ask yourself the same questions.  Will it be any
BETTER in the future?

> Remember back in the 1980's when dozens of companies started making
> 286-based computers? Why is now any different? 

Past experience.  IBM lost the PC market because their senior management
screwed up.  Ken Olsen was right with them saying "the PC is a toy."  
Grove and Intel's senior leadership are different -- they're smart, very
smart.  

So you're wondering how I got the crazy idea that Intel has more than a remote
chance to win the desktop workstation market away from Sun, HP, IBM, Digital,
etc... ?

Intel is using Hitlerian tactics.  Taking one small market at a time without
spooking the whole industry.  If you know what prevented Hitler from taking
all of Europe before the US got into the war, then you'll what Intel's
weakness is.  

[aside:  one thing I've learned is that the computer biz is economic warfare,
plain and simple.  The US Gubmit, WTO, FTC, SEC, etc. impose more rules on
"business" than the UN imposes on the sovereignty of nations...  so we have to
be more clever about it.  The Japanese have been operating their business in
this way since the 2nd World War.  If you're smart, then you can win by the
established rules.]

Intel's Hitlerian tactics are either by design or by effect.  And they're
actually doing it better than Hitler did ... by 1) using less grand-standing,
2) playing more-or-less by established rules, and 3) by not completely
overwhelming the entire markets of their conquests... they leave some
survivors so their opponents can't claim malicious intent.  Reminds me of the
tactics of the US/Soviet Cold War.

> What if Intel refused to sell the CPU at a fair market price -- or refuse to
> sell it at all -- sure they could beat everybody down.  But that would be
> illegal.

Would it be illegal if Intel decided tomorrow not to sell their CPUs anymore?
That would be stupid, not illegal.  Why should they slit their own throats?

But take notice that Intel already sells their chips at a much higher than
average market price for silicon.  Here's some data:  Intel makes ~$40,000 per
8-inch wafer (100 chips * ~$400/chip).  The rest of the industry gets ~$10,000
per 8-inch wafer (400 chips $ ~$25/chip).  Yet the FTC is not stopping them.
Why not?  Because Intel earned the right to do so by playing by established
rules.

In my note in -.1, I proposed the idea that that Intel plans to take over the
PC market because they envy the margins of Workstation makers.  I didn't say
how they'd do it though.  Below are some of my thoughts on the details of
Intel's strategy and how it relates to Hitlerian/Cold-Warian tactics.  I'll
start with chipsets and then go to motherboards.  Then I'll leave you with the
fact that Intel has already announced plans to enter the PC business and leave
the foregone conclusion as an exercise to the reader :-)

Okay, let's look what Intel did to the PC chipset market.  First, to describe
a chipset.  The chipset is the glue that binds the PC together.  It goes
between the CPU, memory, disks, I/O devices, and all of the rest of the
primary PC components and makes sure that it all runs together.  You can't
have a PC without a chipset.  

Intel has used 1) their branding campaign and 2) FUD to win the Pentium
chipset market completely, and do so within one CMOS technology generation
(i.e. < 2 years).  How?  With a one-two punch.  

The Left, Intel is the only chipset maker who owns its own FABs.  What does
this mean?  It means that they can use FUD to scare their customers into
dropping their competition.  AMI (and Opti, among others) have to buy FAB
space from LSI logic and/or VLSI Technology Inc. (VTI) or others ... but there
are 10,000 other chip design houses out there willing to take lower profit
margins to give LSI and VTI higher profit margins -- net result...
Opti's/AMI's wafers are not guaranteed beyond the "good will" of LSI, VTI,
etc. to honor their contracts -- but what happens when the contracts come up
for renewal?  Therefore, AMI/Opti/etc. can lose most or all of their
manufacturing capacity without warning.  Intel used the worldwide shortage in
FAB capacity to leverage this FUD.  [The irony of the situation is that
Intel's own motherboard business is ruled by their own FUD -- and we (DS) have
used this as a lever to sell them our video chip products (since we own our
own FABs).]

And the Right, Intel's branding campaign has given them a psychological
advantage in the market.  PC makers will chose motherboards containing Intel
chipsets because that's what 1) they think to be the best and that's what the
system customers (you and me) are asking for.  Intel produced the Triton
chipset.  It has two unique attributes: 1) it had superior performance to all
other Pentium chipsets at the time of its availability, and 2) it was first to
market.  Thus, it successfully eliminated all competition.

And when AMI and Opti go bleeding to the FTC, the FTC tells them: "Tough sh*t,
pal, Intel clobbered you fair and square."  Now, this of course assumes that
the Intel Neptune/Triton/Orion chipset people didn't have an unfair advantage
in knowing Pentium innards before AMI/Opti/etc.  Truth is, that the internal
people ALWAYS get the first crack at the data.  But that's another essay.


Changing gears...

Now let's take a look at PC motherboards.  What fraction of Pentium
motherboards are made by Intel?  About 90+%.  Sounds like more monopoly to me,
eh?  What's keeping them from dropping their prices and killing off Micronics
and ASUS?  Simple: they don't have to.  They were first.  They won.  But how
did they get there so fast?  Blitzkreig.  Hit first, hit hard, move fast.

Their motherboard people have the same unfair competitive advantage that the
chipset folks do -- early access to Pentium data just like the chipset folks.
When the chipset folks have a timing problem, the motherboard people are the
first to learn about it so they can compensate for the problem and offer their
product before the competition.  They were first to market with Pentium
motherboards, therefore they won.

And second, Intel has been very successful in selling the "one stop shop"
concept to their system partners (Gateway 2000, etc.).  One part number on a
ledger (Intel motherboard w/ cpu) has half the "paperwork cost" of two
(motherboard from ASUS, CPU from Intel).  Call it "Value added," call it
whatever you like.  People prefer the simpler solution and are willing to pay
a premium for it.  So why would anyone ever want another motherboard?

Third, most if not all Pentium PC motherboards other than Intel (and a few
other folks like Compaq and Digital) are made overseas.  So the FTC will be a
little lax in policing Intel's unfair trade practices if it merely leads to
the demise of _offshore_ competition.  If Compaq chooses to junk it's
motherboard business, the FTC will probably not do anything as long as the
stability of the Compaq and/or Digital _corporations_ are not threatened.

But now ask Ted Waitt what he thought when he read the recent WSJ
techno-blurble about Intel's plans to enter the PC systems/servers business.
He's probably not happy.  Why not?  What _could_ Intel do that would put him
and all the other PC makers out of business?  And what gives Intel this
capability? 

And finally, the e-mail response said:

> If you build it, they will come.

Ahhh... but how Digital has learned that fallacy.  Should be more like:

    If they want it, and you can build it at the right price, then they'll buy
    it from you.

Understanding the first part ... "if they want it" is the key to marketing and
more importantly understanding and executing:

                 * * * * * * * * * * * * * * * * * * 
                 * * *  A D V E R T I S I N G  * * *
                 * * * * * * * * * * * * * * * * * * 

It's something that Digital knows very little about, conversely Intel and
MicroSquash have a very keen understanding of this.  Hence, why I feel that
they will eventually win in the computer industry.

The second and third parts are simple, but contain some subtleties... the
difficulty of competing with a monopoly goes as the size of that monopoly.
And Intel will grow to a point where they have enough capacity (and the
efficiencies and economies of scale) that it doesn't make "good business
sense" for anyone else to enter the market.  Thus, nobody else will be able to
build it at the "right price."

Every business day, Intel grows by ~$100 million.  Having seen what they've
done in the PC market, it doesn't take too much imagination to see that they
_can_ do the same in the workstation market.  They already own the bulk of the
software ports.  

They already have a fast OS-on-chip solution: NT-on-P6.

All they need is a reasonable motherboard and *BOOM*:  Intel Workstations.

How long do we have to wait until we realize that Intel will beat us in the
server/workstation markets unless we start a *VERY* *BIG* NT-on-Alpha branding
ADVERTISING campaign?  (Hint: Advertise "NT-on-Alpha" to the public).

How long do we have to keep the strangle hold on Digital Semiconductor's
process development groups until we realize that Intel's process technology
can give them faster chips?  

How long do we have to keep the inefficient workstation design cycles that
cannot offer a new system within two weeks of the day a new microprocessor
reaches revenue-ship?



Bob Palmer, Ed Caldwell, Enrico Pesatori ... are you listening?  Do you agree
with my hypothesis?  If not, then consider the _hypothethical_ case that you
do.  

A very successful Senior Executive Vice President of Sales once told me
"Executives are paid to solve problems and grow profits."  Yet you have
problems, and you're not yet making enough profits (compared to the rest of
the industry).  

If you don't fix these problems and fix them soon, Intel will take what they
are learning in the PC market and crush us.  Then Digital will make no money
and your problems will be of such magnitude that you cannot control them.  Do
you want to live with such a failure?  Will you be able to sleep at night
knowthing that you _could_ have avoided it?  What are you doing to fix your
problems so Digital can make more money?

This notesfile is the focus of the collective consciousness of Digital
Equipment Corporation.  Bob Palmer, your notes to this notesfile are a good
first step, and a good example to the rest of Digital's senior management.  I
can only hope that they follow your lead ... and listen to the responses they
get.  


- jeff

    Intel's strategy has certainly become own the microprocessor market,
    then own the market for motherboard then own the PC market. They are
    on track to do this by the end of the decade.  Intel produces amazing
    amount of PC.  Most of these PC's have other vendors names on them.   A
    plant in Dublin, Ireland produces PC's for Gateway, Dell and others
    for sale in Europe. The customer places and order with lets say Gateway,
    which is forwarded to Intel.  Intel builds the system, puts the
    a Gateway case with logo on the system, put is in a Cow-box with Gateway 
    docs and ships it to the customer.  Gateway never touches the machine.
    This is the future of the PC market.  

    Andy Grove can say there will only be two processors at the end of the
    decade, because market share wise this will probably be true.  Intel will
    sell say 400 Million, The Power people will sell 100 Million and the
    all the others including Alpha will just look like static.  
     
    -Bruce

    Re: last few
    
    Don't agree on PowerPC.  It's reputed to be a poor design with severe
    performance constraints.  The 604 is a disappointment, the 620 can't
    get out of its own way and needs a redesign that may render it
    incompatible with previous models -- Apple's included.  The PowerPC has
    reached this far because of the tremendous clout of three of the biggest
    names in the business, but lagging performance should drive it down 
    within five years, probably less.
    
    As for Intel's future dominance, the reasoning is OK, but it won't
    happen, at least not for long.  Why?  Because it never has.  IBM was 
    seen as "dominant" for a while, but no longer.  In the 1980s, Digital
    thought Digital would become dominant -- but no longer.  Neither Andy 
    Grove nor Bill Gates is either infallible or immortal.  No individual, 
    company, or government has ever dominated everything forever.  If that 
    were possible, we would all be hailing Caesar or swearing allegiance to 
    some descendant of the Khans.  And if the Sunnyvale-Redmond Axis is ever
    perceived as a threat to free enterprise, Congress will bust it up.
    Bet on it.
    
    M.
    
    PS:  The Third Reich was predicated primarily on personal loyalty to
    Hitler.  Even if it had managed to win World War II in Europe militarily, 
    it would have collapsed violently upon his death or even before, because 
    of its total internal corruption and motivation by common hatred.

How dependent is Intel on Microsoft staying ahead (or just a bit behind) in
the applications/functionality space?  Do they have any inherent barriers
(in terms of planned chip designs) to jumping on another software bandwagon
if one starts up?

- paul

    Some Wall Street analysts do not share some of rosey views given about
    Intel in the previous replies.  Many feel that Intel is setting itself
    up for some big trouble.  The chip business has historically a
    boom/bust business.  This does not mean it will continue to be, just
    that it has up to this point in time.  Intel is adding tremendous
    manufacturing capacity.  If the "bust" (could be in the form of a
    recession) comes, as many predict, then Intel is going to be stuck with
    a lot of over-capacity, which will result in them racking up some big
    losses and their stock taking a big hit.
    
    I am not saying this will happen, just that some on Wall Street think
    it will and would not touch Intel's stock with a ten foot pole.
    

    Yep, I agree with .89. Expanding and making big noises is the easy bit.
    But as we know at Digital, it is retreating and keeping things tight
    which is the hard bit. I wonder what their balance sheet is like. Ever
    since I was working for a company that got taken over by Robert Maxwell
    using Junk bonds I have been suspicious of big spenders.
    
    And of course, I agree with the Mandy Rice-Davies quote in .1. 
    
    Kevin

I honestly have NO idea what Intel's Balance Sheet looks like, and my 
"guess" will prbably be shortly corrected, but I suspect the have
VERY little (if any) LT Debt, and Huge equity.

They are a very financially sound company. That is not to say, that if they 
stumble, or there are some other fundamental changes in the supply/demand 
situation fort semiconductors, that these major investments they're making
won't come back to haunt them.

I seem to recall that their capital spending plans run about $3.5 - $4.0B 
per year...

SteveS

(can't recall if this has already been mentioned, but...) I wonder how much
of Intel's success to date can be attributed to the continued success of
Microsoft (one platform, and all that), and if so, how that will continue
now that NT is being sold as a more or less platform independant solution.
This will become more obvious if software vendors allow a free (or at least
low cost) exchange of software which has been ported to the other platforms...

Chris.

re: .-1

Even worse, what happens to Intel if Microsoft suddenly falls behind the
pack software-wise (like what if the BeBox really takes off?).  I know
it's a pretty unlikely possibility now, but then IBM was invincible once
upon a time too...

- paul

    re: .-1 and "What ifs"
    
    What if Microsoft *does* fall behind?  Since we've bet a lot of Alpha's
    future on Microsoft and Window-NT, then Digital still loses out. 
    
    You can speculate about giants falling all you want, but wishful
    thinking is no substitute for a well-reasoned business strategy.
    The one thing that we *can* do is market what we've got, and what
    we're gonna have. Whereas Intel pre-announces their processors and
    design strategies years in advance, we hide products all but ready
    to ship under non-disclosures. Whereas Intel as the market leader
    can afford some negative press (like the Pentium screwup) and survive,
    we as the also-ran must be willing to go the extra mile to avoid any
    negative impressions. We must be helpful, humble, and eager to do
    business without running our customers through scads of red tape
    and "Digital Internal" non-value-added exercises. Only then do we
    have an even chance to survive the processor wars.
    
    Geoff

There is a load of financial guff on Intel at this site, including the balance
sheet.

http://www.intel.com/intel/finance/annual/fnotes/index.html

I have not had time to look at this in detail and would not claim the
experience of such matters to make a well informed judgement, but it seems to me
that the cash position is not enough to finance the investment and that they
are borrowing to finance their investment program. If I understand and remember
it right, Digital had 6* the amount of cash reserves that Intel has now in 91.
So Mr Grove may be doing his favourite halloween number or he may be whistling
in the dark, thinking about what a hugh gamble Intel is taking. As I write this
I have a memory of a certain Mr Fallotti going on about what you can do with 6
billion dollars, back in 92.

I would not invest my pension savings in them on a 7 year basis, but maybe
someone with more experience of corporate accounting  could look at it?  

Kevin

This is the balance sheet



           Consolidated balance sheets 

       December 31, 1994 and December 25, 1993 

(In millions--except per share amounts)                    1994         1993
-----------------------------------------------------------------------------
Assets
Current assets:
     Cash and cash equivalents                           $1,180       $1,659
     Short-term investments                               1,230        1,477
     Accounts receivable, net of allowance for
       doubtful accounts of $32 ($22 in 1993)             1,978        1,448
     Inventories                                          1,169          838
     Deferred tax assets                                    552          310
     Other current assets                                    58           70
                                                       ---------    ---------
Total current assets                                      6,167        5,802
                                                       ---------    ---------
Property, plant and equipment:
     Land and buildings                                   2,292        1,848
     Machinery and equipment                              5,374        4,148
     Construction in progress                               850          317
                                                       ---------    ---------
                                                          8,516        6,313
     Less accumulated depreciation                        3,149        2,317
                                                       ---------    ---------
Property, plant and equipment, net                        5,367        3,996
                                                       ---------    ---------
Long-term investments                                     2,127        1,416
Other assets                                                155          130
                                                       ---------    ---------
     Total assets                                       $13,816      $11,344
                                                       =========    =========
Liabilities and stockholders' equity
Current liabilities:
     Short-term debt                                       $517         $399
     Long-term debt redeemable within one year               --           98
     Accounts payable                                       575          427
     Deferred income on shipments to distributors           269          200
     Accrued compensation and benefits                      588          544
     Other accrued liabilities                              646          374
     Income taxes payable                                   429          391
                                                       ---------    ---------
Total current liabilities                                 3,024        2,433
                                                       ---------    ---------
Long-term debt                                              392          426
Deferred tax liabilities                                    389          297
Put warrants                                                744          688
Commitments and contingencies
Stockholders' equity:
     Preferred Stock, $.001 par value, 
       50 shares authorized; none issued                     --           --
     Common Stock, $.001 par value, 
       1,400 shares authorized; 413 issued
       and outstanding in 1994 (418 in 1993)
       and capital in excess of par value                2,306        2,194
     Retained earnings                                    6,961        5,306
                                                       ---------    ---------
Total stockholders' equity                                 9267        7,500
                                                       ---------    ---------
     Total liabilities and stockholders' equity         $13,816      $11,344
                             

                  I N T E R O F F I C E   M E M O R A N D U M

                                        Date:     08-Nov-1995 08:48am EST
                                        From:     HENDRICKSON
                                                  HENDRICKSON@USCTR1@MRGATE@USCTR1@MRO
                                        Dept:      
                                        Tel No:    

TO: See Below
 
Subject:  

        +---------------------------+  TM 
        |   |   |   |   |   |   |   | 
        | d | i | g | i | t | a | l |      INTEROFFICE MEMORANDUM 
        |   |   |   |   |   |   |   | 
        +---------------------------+ 
 
        TO:     Marketing Council       DATE:   03-November-1995 
                                        FROM:   Bob Hendrickson  
        CC:                             DEPT:   SBU Competitive Marketing 
                                        EXT:    297-3860 
                                        LOC/MS: MRO1-3/F1 
                                        ENET:   USCTR1::HENDRICKSON 

        SUBJECT:   Intel's Pentium Pro Announcement - Nov. 1, 1995

What Intel announced: 
---------------------	
		Intel announced the new P6 family of processors (up to 200 
	Mhz) at its Pentium Pro kickoff event in San Francisco. The 150 
	Mhz and 180 Mhz will be in production quantities in Q4'95. The 200 
	Mhz version, originally scheduled to begin shipping in the 
	June/July'96 timeframe, is now 	scheduled for production quantities 
	in Q4'95 and will likely appear in systems by Q1'96. 

     ------------------------------------------------------------------
     Processor         MHz   SPECint92  SPECfp92  Cache  Avail.  Price
     -------------------------------------------------------------------
     Pentium Pro       200     366.0     283.0   (256KB)   NOW   $1,325  
     Pentium Pro       180     327.0     254.0   (256KB)   NOW   $1,075 
     Pentium Pro       150     276.0     220.0   (256KB)   NOW	 $  974

     Pentium Pro *     166     327.0     261.0   (512KB)   Q196  $1,682 
     Pentium Pro *     200      n/a       n/a    (512KB)   Q296  $1,989 


What Digital's PCBU announced:
------------------------------
		The PCBU announced the Celebris XL 6150, with Intel's 
	Pentium Pro 150 Mhz processor. Current Pentium based Celebris XL 
	systems can be upgraded via a Pentium Pro daughter card. Pentium 
	Pro processors will be incorporated in the Prioris PC Servers in 
	early 1996. The PCBU is currently demonstrating a 200 Mhz Celebris 
	XL.
		PRICING:	
			1) Minimum system configuration - $4,599
			   includes: 16MB RAM, 256KB cache, floppy,
			             Keyboard, and mouse.

			2) Midrange system configuration - $5,549
			   includes: minimun above plus 4X SCSI CD-ROM, 
				     Matrox 3D graphics, 1GB SCSI disk,
				     and Windows NT (factory installed).

			3) High-end system configuration - $9,849
			   includes: 32MB RAM, 4X CD-ROM, 2GB SCSI disk,
				     AccelGraphics AG300 3D Adapter,
				     UltraSCSI adapter, and Windows NT.

			4) 150 Mhz PowerGrade for existing Celebris XL
			   users is priced at $1,999

ALPHA & PENTIUM PRO processor Performance Comparison:
-----------------------------------------------------
	*********** ALPHA ***********    ******* Pentium Pro ********
	Rev/MHZ	  SPECint92  SPECfp92	  MHZ	SPECint92   SPECfp92
        -------	  ---------  --------	 -----	---------   --------

	4/233	    157        184	   ---      ---	      ---
	4/266	    198        263	   ---	    ---	      ---  
	4/275*	    202.9      293  	   133(NA) >200	      ---
	5/250*	    277	       410 	   150	    276**      220         
	5/266	    289	       405         166	    327        261
	5/300	    338	       502	   180	    327        254        
	5/333	    400***     ---   	   200	    366        283 
	5/417       500***     ---	   ---      ---        ---
         
   * = NOT avail in  Workstations,  NA = never annc't,   *** = estimated

       ** The Celebris XL 6150 with a Pentium Pro 150 Mhz processor has the 
	  equivalent SPECint performance of an Alpha EV5/250 processor,
	  however, since the ALPHA 5/250 is a variant of the Alpha 5/266,  
	  the next comparable chip is the EV5/266.  

	Intel's Pentium Pro processors (150 Mhz - 200 Mhz) performance in
	SPECint'92 shows relative positioning against the Alpha EV5 (250 
	Mhz - 333 Mhz), however Alpha has a 2 to 1 advantage in SPECfp.

Intel's Pentium Pro Background:
-------------------------------
		In September, Intel bypassed the planned 133 Mhz version 
	of the Pentium Pro in favor of the better performance of the 150 
	Mhz on desktop systems. Both the 166 Mhz and 200 Mhz Pentium Pro 
	Chips will have bus speeds of 66 Mhz versus the 60 Mhz bus speeds 
	of the 150 Mhz and 180 Mhz chips. The faster bus speed (66 Mhz) of 
	the 166 Mhz and 200 Mhz chips suggest that they are better suited 
	for the server market.
     							        
	Table 1:	CLOCK	      Bus Speed		     Target Market
			-----	      ---------              -------------
	  (unannounced)	133 Mhz	    66 Mhz (2.0 x 66 = 132)	Server
			150 Mhz	    60 Mhz (2.5 x 60 = 150)  Desktop/Wrksta
			166 Mhz	    66 Mhz (2.5 x 66 = 165)	Server
			180 Mhz	    60 Mhz (3.0 x 60 = 180)  Workstation
			200 Mhz	    66 Mhz (3.0 x 66 = 198)  Wrksta/Server

	Furthermore, Intel announced two chipsets; the 82450GX (Orion) and 
	the 82450KX (Mars) . Originally the "Orion" chipset with its 
	increased functionally, 4-processor support, and higher price was 
	scheduled for the server market. The "Mars" chipset with 
	2-processor support and smaller footprint (half the size of Orion) 
	was designed for the desktop market. Problems with 4-way SMP 
	functionality in the Orion and 	delays in the development of "Mars"
	forced Intel to debut a variant named "Orion DT" (DT for  DeskTop) 
	with 2 processor support at the larger footprint. The true Mars 
	chipset will ship in Q1 CY96 along with the 4-way SMP Orion. 

     		Orion                             Mars
		-----				  ----
	     * Due Q4 1995 (now due Q196)      * Due Q1 1996
	     * Designed for servers            * Designed for desktops
	     * Supports four processors        * Supports two processors
	     * Supports two PCI buses          * Supports one PCI bus
	     * Large and expensive             * Small and inexpensive
  
The Competition:
----------------
	   Vendor	Model 	    Price 	Configuration

	   HP         Vectra XU    $5,700   PP150 CPU, 16MB RAM, 1.0GB disk
	   NEC	   PowerMate 150  >$5,000   PP150 CPU, 16MB RAM, 1.6GB disk
	   IBM        PC360       $4K-$5K   PP150 CPU, 32MB RAM, 1.2GB disk
	  Compaq   "Super Client"   ----    PP166 CPU, 16MB RAM, 1.0GB disk            
	Intergraph    TDZ-600	    ----    a quad processor 3-D graphics 
         				    system that will compete 
					    against SGI's Indy & Indigo2. 
   ** Traditional PC focused players such as Dell & Gateway will now have        
      P6-based Personal Workstation offerings. 

Analysis:
---------
		With the announcement of the 200 Mhz Pentium Pro processor, 
	Intel has fired yet another round in the greatly debated Intel 
	versus RISC war. The immediate impact of this announcement will be 
	the push of high end Intel-based desktop products into the 
	traditional Unix workstation market. In Q1'96 when the 4-way SMP 
	PCI chipset is scheduled to ship, Intel-based server performance 
	will be greatly enhanced due to the good SPECint performance of the 
	Pentium Pro.

		The SPECint performance of the 200 Mhz Pentium Pro warrants
 	that at least an EV5 version of the Alpha chip is required to 
	compete (most likely the EV5/333), however, it is the 150Mhz 
	Pentium Pro being delivered in quantity today. The 200 Mhz version 
	was announced but is not volume shipping and 200 Mhz based systems 
	aren't due out until Q1'96. Also, the SPECfp performance (which  is 
	the measurement of performance in compute intensive, rendering, and 
	hi-end 3D graphics applications) is sorely lacking. The Pentium 
	Pro announcement directly affects the positioning of the low end 
	workstations such as the AlphaStation 200 and 250 lines as well as 
	the Alpha XL Personal Workstation systems.    

		Vendors such as IBM, HP, Compaq, and Intergraph are poised 
	to exploit the performance of the Pentium Pro family in their own 
	respective versions of a "Personal Workstation" with entry pricing
	in the $4K to $5K range and a full range of available graphics 
	accelerators. They will compete directly with the Alpha XL systems
	since the announcement of the Pentium Pro has temporarily narrowed 
	the performance differentiator that the Alpha architecture brought 
	to Digital's "Personal Workstation" family.
		
		Most analysts agree that the first systems will be sold to 
	power users and developers since true 32 bit applications are not
	yet readily available to exploit the full potential of 	Intel's 
	Pentium Pro. Remember that the Pentium Pro does not show any 
	performance gains on 16 bit applications and many 32 bit operating 
	systems and applications still have embedded 16 bit code. However, 
	it is difficult to believe that the availability of true 32 bit 
	applications that will capitalize on the performance capability of 
	the Pentium Pro will be long awaited.

		The question to Digital is how fast do we upgrade the 
	low-end workstation and Alpha XL families with faster Alpha chips 
	in order to maintain performance superiority to continue to justify 
	the Alpha performance premium. The performance issue is confusing 
	since Pentium Pro shows little difference between Integer and 
	floating point performance, whereas, a comparable Alpha chip with 
	the same integer performance as a P6 has approximately twice the 
	floating point performance. The main opportunity for the Pentium 
	Pro is to push high-end desktops into the Unix workstation market 
	where high-end 3D graphics, rendering, and compute intensive 
	applications are normally used. Alpha with its superior floating 
	point performance still has the advantage here. 

Summary:
--------
		The Pentium Pro processor has raised the performance bar 
	of the desktop systems to a level that allows them to compete in 
	the low-end to mid-range Unix Workstation market as well as in the 
	Windows NT Personal Workstation arenas. To maintain the Alpha 
	performance "differentiator" in these systems, Alpha microprocessor 
	upgrades to EV5 are essential. Intel server systems are not 
	expected to see the Pentium Pro until Q1'96 due to SMP issues on 
	the Orion chipset. 
		Companies such as HP, IBM, Compaq, and Intergraph are all 
	poised to take immediate advantage of the Pentium Pro from a 
	hardware offering perspective and will offer Pentium Pro Personal 
	Workstations (Hi-end desktop offerings with full line of graphics 
	capabilities). To take full advantage of the Pentium Pro, true 32 
	bit operating systems and applications software still need to be 
	developed and/or ported. Q2'96 is the current estimate for P6 on 
	corporate desktops but I believe Compaq and Intergraph will push 
	strongly to accelerate this schedule. 

		HP and IBM will carefully position their new Personal 
	Workstation lines so as not to severely impact their traditional 
	low-end workstation offerings. HP will have to be more aggressive 
	to solidify and support its story around its HP/Intel strategy and 
	P7 roadmap. Compaq, Dell, and Intergraph with "Intel only" product 
	lines will focus sharply on getting essential applications 
	developed for these hi-end desktops in an attempt to capture 
	marketshare in the Unix workstation market. Compaq has stated it 
	will concentrate on the development of the "Super Client" which 
	will be a DeskPro variant. Also expect Compaq to push the 
	price/performance message that has worked so well in their server 
	marketing strategy. Intergraph, already rated #2 in the GIS 
	marketplace in PC workstations, has a full line of graphics 
	accelerators including hi-end 3D and will probably become a very 
	prominent competitor. Intergraph has already announced its 
	intentions to attack SGI's INDY and INDIGO 2 workstations with a 
	4-way version of the Pentium Pro Personal Workstation, the TDZ-600.  


		The Intel announcement appears to spell trouble for Sun who 
	has no NT startegy and weak UltraSPARC-I performance ( avail. in 
	Q1, the UltraSPARC-I chip @ 182 Mhz has 260 SPECint92 and 410 
	SPECfp92). Intel's Pentium Pro @ 200 Mhz (366 SPECint92) readily 
	beats the UltraSPARC-I. Sun is porting Solaris 2.5 to Intel with no 
	current thoughts of supporting NT on the SPARC architecture.

		SGI is also on precarious ground as they also do not have a 
	formalized strategy to support NT. The MIPS technology does 
	support NT but SGI has publicly denied the existence of NT. 
	Intergraph is using the Pentium Pro to formally target SGI's MIPS 
	based workstations.

		Today, HP is limited to Windows NT on their Intel platform
	only. PA-RISC does not support NT (don't be fooled by HP's 
	bi-endian story line around PA-RISC) and HP does not intend to do 
	so until P7 which is 18-24 months away and then it will probably 
	require a migration on the end-user's part.   

		Digital is in a very enviable position as we support
	NT on both our RISC and CISC platforms. Digital offers a full line 
	of workstation products from Alpha XL Personal Workstations to 
	sophisticated high-end 3D graphics workstations and they all 
	support Windows NT. The Personal workstations can support either  
	Intel (including P6) and Alpha processors based on the customer's. 
	requirements. Digital's current Alpha offering outperforms Intel's 
	current Pentium Pro offering. Alpha processors offer TWICE the 
	floating point performance of a Pentium Pro. Intel's next 
	performance and frequency thrust will probably be from the 
	future generation P7 (12-18 months away and a required migration).
	Alpha offers investment protection through simple board upgrades 
	with no migation necessary unlike Pentium to Pentium Pro upgrades.  
		
	** Digital needs to articulate an overall Intel/Alpha product 
	strategy that can be easily understood both internally and 
	externally to Digital, especially to its sales force and customers.

    The P6 analysis focuses on hardware, but I have questions about the
    software side as well:
    
    It will be interesting to see the cost comparisons between Alpha and
    Pentium Pro *solutions*. Not just hardware, but operating systems,
    network, and user applications as well. How will Alpha compete?
    
    Of course, the Intel/NT platform will have %100 applications share
    right from the start. Anyone who develops NT applications will do
    them on the Intel architecture, then port to Alpha, right?
    
    Finally, the age-old question of floating-point performance: does
    it really matter to the vast majority of software applications?
    
    Geoff

    Re: .97
    
    If I understand the publically announced stuff right, applications
    for NT Alpha and NT Intel from MIcrosoft will ship at the same time
    and at the same price.  What other application writers will do, I have
    no idea.  Plus, there's FX!32.
    

    re .98: That's what M$ keeps saying, but unfortunately, it's not true.
    E.g. Visual C++ for Alpha costs much more than the Intel version.

    does this one know how to divide correctly?

    
    
    What is the plan to push the faster Alpha's on the desktop at P6 
    200MHz prices?
    
    Kevin
    

    Re: 99 "Visual C++ for Alpha costs much more than the Intel version"
    
    Visual C++, the version due out the end of this month, will cost
    the same on both platforms.  Also, a number of Microsoft NT
    apps are now shipping with both Intel and Alpha versions in the same
    package.
    

    re .102: ok, I stand corrected then... but it used to cost more, didn't
    it?
    

  And then there's "applications" versus "applications". Exactly
  how much of their catalog did Microsoft mean when they said that
  applications would be available on Alpha?

  Did it include the entire "Microsoft Home" set of CDs, growing
  by leaps and bounds every day? Did it include the more obscure
  business products like "Microsoft Project"? Did it include
  the client to access the Microsoft Network? Microsoft "money"
  and whatever follows that? Etc...

                                   Atlant

>    Visual C++, the version due out the end of this month, will cost
>    the same on both platforms.  Also, a number of Microsoft NT

     I got mail today saying VC++ 4.0 for Alpha has been released
     to manufacturing! Coming soon to a theater near you.

    Re: .104  how much of their catalog did Microsoft mean when they said
    that applications would be available on Alpha?
    
    The wording I saw was that all the apps Microsoft produces for
    Intel NT will be available for Alpha NT, and ship at the same
    time.  I didn't see the equal price mentioned in the original
    annnouncement, but Dec has clearly worked this issue.  This
    is one thing that seems to be being done right.
    

Not only same price, but same functionality!

		Steve

> The wording I saw was that all the apps Microsoft produces for
> Intel NT will be available for Alpha NT, and ship at the same
> time.

  This still allows a tremendous "shading" of the answer.

                                   Atlant

>    Visual C++, the version due out the end of this month, will cost
>    the same on both platforms.  Also, a number of Microsoft NT
>    apps are now shipping with both Intel and Alpha versions in the same
>    package

I find this very encouraging. Lurking at the back of my mind was always
the suspicion that even if we got MS to ship simultaneously, we may lose
out on the distribution channels.  This way the vendors don't have to
worry about shelf space.

But it sounds a little too good to be true... Have MS favoured us over
MIPS and PowerPC?  How about documentation? do the standard docs
cover both (all?) platforms?

Mark

  

Re .106:

From BP at the stockholder's meeting:

* Microsoft has promised to release all *server* software simultaneously on
Alpha/NT and Intel/NT.

* Microsoft has promised to release all *client* software simultaneously on
Alpha/NT and NT for any other RISC architectures.

In retrospect, this sounds like what I heard from Robert Bismuth's presentation
last month.

I bet the applications run by the average PC user are almost completely "client"
apps.
				/AHM

> * Microsoft has promised to release all *client* software simultaneously
> on Alpha/NT and NT for any other RISC architectures.

  So "Never on Alpha or MIPS or PowerPC" would still satisfy this
  criterion, right?

  Heck of a deal! #&@^!#^$

                                   Atlant

I didn't hear BP at the stockholders meeting, but did hear Bob Bisumuth 
discuss the NT situation at TSU last month. As I recall, what was said re the 
server apps was accurate, but what MSFT has agreed to re client software is 
different. MSFT will not release any of their client apps on any RISC 
platform before Alpha, hence Word can't be released on PPC or MIPS before 
it's released on Alpha. This DOES mean, as the previous noted, that NEVER 
remains a possibility.

SteveS

Re .112:

> what MSFT has agreed to re client software is different. MSFT will not release
>any of their client apps on any RISC platform before Alpha ...

Yes, that's a better phrasing of what I meant to say.
				/AHM

>		In September, Intel bypassed the planned 133 Mhz version 
>	of the Pentium Pro in favor of the better performance of the 150 
>	Mhz on desktop systems.

P6 is a disappointment.  Early performance tests show little or no improvement
over Pentiums for the majority of existing software.  This is due
in part to Intel's decision to optimize for 32-bit server performance
(SPECint and SPECfp) over 16-bit desktop performance.

They had to bypass the 133 Mhz version, it was slower than a 133 Mhz Pentium
running Windows 95.

MARKETING TO THE RESCUE:  Reposition fast clock rate P6's (designed with
the newest process technology) not as the next generation, but as an upward
extension of the Pentium family, the "Pentium Pro".  Sell the sizzle,
fast clock rates and high SPEC numbers.

Can Intel really deliver 200 Mhz CPUs in volume and at low cost on schedule?
That's not as important today as competing for CPU performance mind share.

This is classic Intel marketing.  You don't need the best "device" to win.
Sell a complete product including customer expectations and perception of
safety.  By the time the market figures out they've been manipulated, Intel
will have another solution.

>		With the announcement of the 200 Mhz Pentium Pro processor, 
>	Intel has fired yet another round in the greatly debated Intel 
>	versus RISC war.

True, but the war isn't over.  How will the opposition respond?
		
>		Most analysts agree that the first systems will be sold to 
>	power users and developers since true 32 bit applications are not
>	yet readily available to exploit the full potential of 	Intel's 
>	Pentium Pro. Remember that the Pentium Pro does not show any 
>	performance gains on 16 bit applications and many 32 bit operating 
>	systems and applications still have embedded 16 bit code. However, 
>	it is difficult to believe that the availability of true 32 bit 
>	applications that will capitalize on the performance capability of 
>	the Pentium Pro will be long awaited.

Massaging the issue.  Applications are not the most critical bottleneck.
Much of Windows will remain 16-bit for a long time to come.  There's no quick
way to re-write and test this much code while maintaining backward
compatibility.  Windows NT is years away from being accepted as a mainstream
OS for the desktop.

>		The Pentium Pro processor has raised the performance bar 
>	of the desktop systems to a level that allows them to compete in 
>	the low-end to mid-range Unix Workstation market as well as in the 
>	Windows NT Personal Workstation arenas.

The analisys mixes desktops and servers just as Intel intended.
Server performance has been raised, and clevery blured to look like a
significant desktop advance as well.

- Peter

    I read where Compaq will not use the Intel motherboards in servers.  Intel
    is planning a big push into the motherboard market, which is really
    hurting the Taiwanese companies that make them.  Intel wants to
    "standardize and commoditize" motherboards around a standard they have
    developed.  Compaq refuses to go along because Compaq's motherboard
    design offers greater performance and functionality.  Compaq feels that
    standardization is great on the desktop, but not the server.  Compaq
    feels that it motherboard design is what helps differentiate it from
    other Intel-based server manufacturers.  If everyone begins to use
    Intel motherboards in their servers along with Intel chips then all the
    server manufactures will look alike.  There will not be much of a
    reason to buy one over the other, except price, and price is what the
    commodity business is all about (except of course marketing and
    distribution).  
    
    It will be interesting to see if the Intel-based server market becomes
    just like the PC market (cutthroat, volume based, standardized, low
    margin, etc. - you get the picture).
    
    How will this help/hurt the RISC-based server vendors.

    Ahem,
    
    in the papers yesterday there were a few pieces about how Intel is
    laying off 200 contract employees at its plant in Leixlip, Ireland, due
    to poor results in the last quarter and a worsening cost situation. 
    
    This is highly confusing, considering Mr Grove is going to wipe the
    floor with Digital in the next few years. Would the Grove fans who were
    singing Intels praises around Halloween please explain all this real
    quick, before the wind changes and their faces get stuck?
    
    Kevin  
         

    When Intel posted record results a few weeks ago (about $5b per
    quarter now), they said that there were some disappointments
    in the motherboard business.  Namely, they misjudged inventory
    and got caught having to write off some.  Given that at least
    part of their motherboard manufacturing is done in Ireland, perhaps
    the layoff of contract workers is related to the disappointment
    in motherboards.
    
    Needless to say, the system manufacturers that do their own boards
    weren't terribly broken hearted. ;-)
    
    I wouldn't read too much into somebody laying off a couple hundred
    contract workers.  .02 Kratz 

    re: Intel Layoffs
    
    I certainly wouldn't want to compare Digital's layoff numbers with
    Intel's ...
    
    Geoff

    I see Intel are also advertising the sale of one of their PC assembly
    lines.  The reason given was that they had improved the process and
    did not want to upgrade the hardware.  All you need is a few $100k and
    space to put it and you too can turn out 15k PC's a month!
    
    Nice to see the Internet getting more commercial uses.
    
    Calvin

    re .118
    
    I'm not comparing with Digital. However, hiring 220 employees on
    contract and letting them go a few months later is a waste and a pr
    booboo. It shows
    that the chip business is very volatile and Intel do not have the
    resources to ignore the tidal changes. Given the size of their
    investment program and their policy of achieving competitive prices
    through economies of scale, if they DO hit a bad patch, they will be in
    big trouble with thier cost structure.
    
    So to get back to the base-note, I stll think he was whistling in the
    dark to try and stay cheerful when he thinks about the risks they are
    taking.  

    re: .120 
    
    >So to get back to the base-note, I stll think he was whistling in the
    >dark to try and stay cheerful when he thinks about the risks they are
    >taking.
    
    There's no doubt that Intel is in a risky business, compared to many
    other industries, but then, so are we. You allude to Intel's resources,
    but they are greater than ours or anyone else's in their core business.
    If Intel is hit by a stormy season, you can rest assured that it will
    be a hurricane for the rest of us.
    
    In general, the overall volume of the PC business looks like it might
    be slowing down this year. Since there are many semiconductor companies
    who primarily produce parts for PC's, this is not good news. On the
    other hand, there is an overall surge in the demand for high-end PC's,
    laptops, and servers, where the single-unit chip count is higher, and
    the demand for quality or specialized parts are higher. Those vendors
    who make high-end graphics, SCSI, and network chips are likely to see
    the best gains this year, while the ones who make commodity sound, VGA,
    and controller chips are going to get hurt.
    
    Alpha aside, I think Digital Semiconductor has other products that will
    compete favorably in the new climate, so I'm hoping for a banner year.
    
    Geoff  
    

re Note 4189.121 by ZPOVC::GEOFFREY:

>     If Intel is hit by a stormy season, you can rest assured that it will
>     be a hurricane for the rest of us.
  
        Unless, of course, *we* are the storm!

        Bob

    Well I think the point I am trying to make is that Intel is trying to
    grab the market by making a hugh investment in manufacturing capacity
    and thereby achieving large economies of scale. If for some reason they
    do not achieve the innovation they need to stay in the market fast
    enough (P 7 should be called Mirage, when you get near to it it
    disappears into the distance) and they end up with vast capacity for
    products the market doesn't want, it is going to be a bad hangover.
    In that context hiring 220 contractors and then laying them off is
    smaal beer, sure, but it is a laughable way to approach the problem.
    
    To compare Intel with Digital is not really logical, because our
    strength is not based on capacity, but technological lead. It looks to
    me that with Alpha and VLM we are getting back to the lead we had in
    the 80's with VMS and DECnet, before open systems came along, which is
    just as well as we put a lot of chips down on Alpha.
    
    The partnership approach has given us more stability, but it would be
    the end of the road if Grove did manage to put the squeeze on Alpha. 
    
    But that's not the way things are shaping up is it?
    
    Kevin  
     

    Just some additional pseudo-random thoughts ;^)
    
    Anyone else remember that:
    
    1- 	Intel was near bankruptcy in the early-mid '80,
    	and that IBM basically bailed them out?
    
    2-	That Zilog once had a lock on the 8-bit market, but had its world
    	turned upside down by IBM and the PC?
    
    3-	That Intel went single-source on the x386 because of the amount of 
    	competition in the x286 market?  Both ADM and Harris were licensed
    	286 sites, and both competed well against Intel.  Harris even had
    	a 20MHz 286 chip that was measurably faster than the i386-16 in
    	PC-DOS work (and there even might have been a 24MHz chip)?
    
    4-	That there are only three domestic auto manufacturers, and they
    	make or source everything, but trucking is a whole other show?
     
    
    My points being:
    
    1-	Intel is not above making bad business decisions, or the impact
    	of the market.
    
    2-	That radical change can often be overlooked by the market leader
    	until too late (examples abound on this one - some close to home).
    	**FOR EXAMPLE**
    	WHAT WOULD BE THE IMPACT ON INTEL OF THE SUCCESS OF THE $500-1000
    	"Internet Terminal" DEVICE?  This device would not impact Digital,
    	given that we sell Alpha and Intel workstations only to business
    	now (remember the PCBU announcement), and that the device I speak
    	of will probably not be suitable for business purposes.  For Intel,
    	I believe the problem would be more severe.  It has tied much
    	of it's strategy to the expanding home PC boom.  If inovation
    	results in the evolution of a device for home use that by-passes
    	the Wintel cycle of "fatter H/W => fatter S/W => fatter H/W ...",
    	then a good chunk of the base for sustaining Intel's capital
    	structure begins to erode.
    
    3-	Competition could still be a significant factor for Intel.  If the
    	RISC manufacturers (including us) began to adopt a strategy similar
    	to Intel's where the older chips are sold off at very competitive
    	prices, then they might be more price competitive with the upper
    	end of the Intel line.  The 200Mhz P6 has put a damper on this
    	somewhat, but the logic is still sound, especially for PowerPC,
    	SUN Sparc and MIPS, where there are manufacturing partnerships
    	involved.  Couple this with the renewed strength of the Intel
    	clone makers and you have more competition at the high and 
    	medium levels of Intel's line, where the $$$ are made.  Then
    	add chips suitable for the PDA/internet terminal market (like
    	StrongARM ) and you have increasing pressure at all parts of
    	Intel's market.
    
    4-	There seems to be a general assumption that Alpha is a direct 
    	competitor to ix86 chips.  While this is certainly true in
    	the workstation space, I believe that the issue is less clear-cut
    	in the server space and speciality workstation market (3-D, CAD,
    	imaging / floating point -driven applications).  I like to 
    	draw an analogy here between cars & workstations versus trucks &
       	servers.  While the North American car manufacturers went through
    	rationalization some time ago (the last being the absorbtion of
    	American Motors/Jeep by Chrysler), there appears to be room for
    	more than 3 truck manufacturers.  As well truck  engines, bodies,
    	and chassies don't all come from a single manufacturer, nor have
    	GMC, Ford, and Chrysler put the trucking manufacturers out of
    	business.  Thus I feel that is it indeed valid to ask if Alpha
    	can be boxed out of the mass workstation business, leaving only Intel
    	and PowerPC (and possibly MIPS because of NEC), but this discussion
    	should not apply to servers and specialized workstations.  The
    	questions for Digital therefore become: (1) will Intel succeed
    	at building "truck" engines to the exclusion of the smaller 
    	independents (HP has already bet yes, right?), and (2) should Digital
    	remarket it's Alpha "engines" to other truck manufacturers?
    	On the latter, I have only seen workstation ALpha-based clones
    	not servers (and the workstations are usually high-price designs
    	for the specialized market).
    
    Regards ... charles
    
    

    Growth and success of a cheap Internet terminal market might very well
    impact Digital -- in a positive way, what with today's announcement of
    the StrongARM chip, which can match a Pentium 100 for performance at
    one quarter the price and on less than one watt of power.
    
    M

    
      My personal prediction is that a telephone based 'cheap internet 
    terminal'  will be a flop.
    
      On the other hand, A service linked to the new DSS hardware 
    which could fetch pages via a 56kb broadcast link using some 
    unused transponder  space on the DSS  satellites (or local microwave 
    broadcast) will probably make somebody rich beyond their wildest 
    dreams in a couple of years...
    
    
    							-al

>      My personal prediction is that a telephone based 'cheap internet 
>    terminal'  will be a flop.
    
I think it'll be a great success.  Look at the telephone system; most users 
just want a handset they can use anywhere, and not worry about the running of 
the exchange.  I think that's quite a good analogy; give a person a cheap 
display system, and keep the data elsewhere, where it can be accessed anywhere 
(with the appropriate security, of course), with all the other benefits of 
fault tolerance, backups, etc.

Chris.

    
    	I am with you Al...
    
    	An internet terminal is a joke; just like the Newton. The consumer
    does not need another device to carry around that has access charges
    measured by the minute. I cannot believe the "lemming" behavior of so
    many.
    
    	But a device that "grafs" into my DSS that gives me 56Kbs into
    wherever thru my PC as a two-way pipe. Its a winna, folks. Them
    dishes is where the money is.
    
    
    		the Greyhawk

    Dont sell the internet terminal short in business. I suspect 70+% of
    todays users would be better served by a GUI terminal, running Java
    like aps for simple word processing, spreadsheets, and hooked up to the
    corp database via WEB technology. Note our use of Netscape to front end
    the SAP sales order systems.
    
    and the IT departments that have lived with all the support issues of
    PCs and windows might love it.

I think this is one of those things that is going to be the subject of a
massive amount of debate, with portentous statements and weighty commentary
on both sides, and 5 years from now one side is going to look really stupid.

Having said that, I will take my chances on being on that side.

I think the idea of an Internet terminal is short-sighted, the idea of
techno-dweebs who would never buy one themselves but imagine that they
would be perfect for the great unwashed, and in general will be the
biggest marketing blunder since the Edsel.

I think that this is best expressed by .129, where Ron said:

>    Dont sell the internet terminal short in business. I suspect 70+% of
>    todays users would be better served by a GUI terminal, running Java
>    like aps for simple word processing, spreadsheets, and hooked up to the
>    corp database via WEB technology. 

We have had the capability of building this for a long time: it is either 
an X-terminal or a diskless PC or a Multia.  Some of my customers bought 
diskless PCs for precisely the reason that Ron said: ease of support.  But
they discovered two things that killed the whole idea:

1) people need to customize their systems to make them more productive.  I
   am not talking about bringing in a cool screen-saver, I mean the ability
   to upgrade their systems with a special app that only a very few people
   want but that those people truly do need, or a more recent version of a
   piece of supported software that MIS hasn't gotten around to certifying.

   An example of this is the Sales WorkBench HiNote Ultra PCs that were
   given to the Field people.  Many of us have installed software that is
   not supported by the SWB team yet (Visio, NetScape 2.0, Trumpet 2.0,
   and even Windows 95), just to get our jobs done.  We bought this stuff
   with our own money because we need it for the business purposes of the
   equipment, but the SWB team (who are doing an excellent job) does not
   support the plethora of software we are using.  

   If we had the equivalent of the Internet terminal, which was totally
   outside of our control in terms of products, none of this would have
   been possible, and we would be handicapped in getting our jobs done.

2) the network overhead in truly diskless stations.  Multia does the best
   job of getting around this problem by including a local swapfile, but
   swapping over the network can bring an Ethernet to it's knees.

I think the key word in Ron's analysis is "simple".  Simple word processing,
simple spreadsheets, simple queries of the corporate database are things
that a person does for a while, and then they start to want to do more.
They see something in a magazine, they see something a friend did, they see
something that a competitor did, or they just have this idea of how to do
something spiffy.  But they want to go beyond the basics of a "simple"
word processor, spreadsheet or whatever.

With an Internet terminal you can't.  Java is so limited, the amount of
memory and processor in the Internet terminal is so limited (mostly because
of the $500 price tag, including monitor: have you priced a good monitor
recently? the 14" ones are $250, and the 15" ones start at $350.  Not much
left for a CPU, memory, keyboard, mouse and Ethernet card or modem), and the 
network load over Ethernet when you start swapping those full sized app's
will kill your network.

My final point is to ask anyone who advocates this thing if they would buy
one for themselves.  I bet you will get "Well, no, *I* need a *real* PC
because I am a power user, but all those other people should be satisfied
with an Internet terminal because their needs are so limited".  Sorry, I
don't buy that line.  Their needs are limited because of lack of experience
or training, not true business needs.  Look at how well Microsft Works sold.
It didn't.  People (corporations, SOHO and home use) overwhelmingly chose 
the full featured system over the stripped down toy.  I believe that the
Internet terminal is this years Microsoft Works.

And if people truly want it, sell them a Multia.  It offers the ease of use
and low support costs of an Internet terminal with the full power of a PC.
Ok, so it isn't $500.  It is still the same idea.

FWIW and IMHO.

-- Ken Moreau

    I too am a bit sceptical about the Internet terminal idea.
    
    Apart from reasons already stated, the communications infrastructure
    isn't really there yet, and needs huge investitions. ISDN isn't
    widespread yet in most countries, let alone fast enough mobile
    communications via cellular systems (GSM only does 9600 bps) or
    satellite. So you're bound to the old-fashioned analog telephone
    network with 28.8 kbps (if you're lucky).
    
    I would see more opportunities in the corporate space (because of the
    support costs) - whether it's an X terminal, Internet terminal,
    diskless PC, Multia or whatever is another thing (and the
    communications infrastructure is usually more mature).
    

 <<< Note 4189.130 by ODIXIE::MOREAU "Ken Moreau;Technical Support;Florida" >>>

>> With an Internet terminal you can't.  Java is so limited, the amount of
>> memory and processor in the Internet terminal is so limited (mostly because
>> of the $500 price tag, including monitor: have you priced a good monitor
>> recently? the 14" ones are $250, and the 15" ones start at $350.

Clarification: most people touting the $500 price-tag for the
Internet terminal do not include the monitor; Larry Ellison's
dream-device hooks up to the TV, I believe... 

Ram

> Clarification: most people touting the $500 price-tag for the
> Internet terminal do not include the monitor; Larry Ellison's
> dream-device hooks up to the TV, I believe... 

  Which makes it even more of a joke. Now if it hooks up to an
  HDTV, then maybe I'd become a semi-believer, but at *THAT*
  price-point, why not just include an appropriate monitor?

                                   Atlant

    Yeah - there is some creative market demand generation going on here.
    
    These are being referred to as "set-top" boxes I believe.  This market
    will only take off if they can package it through a creative mix of
    what are currently quite separate selling channels ;
    
    		- home/business software
    		- whitegoods/browngoods
    		- cable/satellite/subscription TV
    		- network services (phone co. + data networks)
    
    Now the person that cooks that bunch up and offers (at the point of
    sale of the TV), a package that turns it into an Internet browser with
    pre-paid/guaranteed useful contents all accessible with the remote
    control will have solved the purchasing argument of ;
     
    	"Is it really worth an extra 500 bucks?" 

    
    the only people who showed up at the internet expo was sun with a
    protype set-top box for JAVA.
    
    bob cansler
    

    Don't internet terminals exist today under the guise of used 386 and
    older 486 PCs?
    
                                            August G. Reinig

    Puts Supercomputing in the Palm of your Hand ....
    
             DIGITAL TARGETS SUPERCHARGED StrongARM CHIP
                   AT CONSUMER ELECTRONICS MARKET
    
     ....Enables Better Handwriting Recognition and True Speech
      Recognition for PDAs, Sizzling 3D Games, Interactive Home
                  Shopping, Fast Web Terminals....
    
    
        MAYNARD, Mass., February 5, 1996 -- Digital Equipment
    Corporation today introduced the much anticipated SA-110
    StrongARM microprocessor, the first processor to combine the
    performance of a supercomputer with power dissipation low
    enough to run on AA batteries, and pricing which is geared
    toward mass-market, consumer electronics products.
        "The StrongARM microprocessor family is one of the
    cornerstones of our merchant vendor strategy," said Ed
    Caldwell, vice president, Digital Semiconductor, a business
    of Digital Equipment Corporation.  "We see tremendous
    opportunity to deploy this technology across many mass-
    market application areas."  According to industry analysts,
    the potential microprocessor volume for the StrongARM target
    markets -- personal digital assistants (PDAs), electronic
    organizers, set-top boxes, and video games -- will exceed 29
    billion units by 1999.
        The SA-110 microprocessor is the first member of the
    StrongARM family resulting from the architecture license
    agreement between Digital and Advanced RISC Machines Ltd.
    (ARM), developer of the ARM 32-bit RISC architecture.
    "Combining ARM's low-power architecture with Digital's high-
    performance processor design expertise and CMOS process
    leadership has created a new paradigm for embedded consumer
    electronics products -- supercomputer class performance on
    AA batteries," said Robin Saxby, president and CEO, 
    Advanced RISC Machines Ltd.
        
    MARKET OPPORTUNITIES
        
        Industry watchers see a bright future for the StrongARM
    technology.  According to Jim Turley, senior analyst,
    Microprocessor Report, "The SA-110 StrongARM processor is a
    technical tour de force and a milestone for both Digital
    Semiconductor and the ARM architecture.  It offers a nearly
    unbeatable combination of performance, price and power
    consumption."
        Tim Bajarin, president, Creative Strategies
    International commented, "The design of the SA-110 StrongARM
    chip has clearly involved PDA developers, smart phone
    manufacturers, set-top box suppliers, and even companies
    exploring the internet computer.  This type of foresight
    heralds a new wave of mobile products which meet consumers'
    real needs."
        
    INTERNET TERMINALS
        
        The huge market potential for an instant access, low-
    cost product to 'surf the net' has grabbed the attention of
    both the computer and consumer electronics industries.  The
    SA-110 StrongARM chip is well-positioned to make this
    concept a reality.
        According to Andy Laursen, vice president of Network
    Computing at Oracle Corporation, "The explosive consumer
    demand for inexpensive, high-performance internet access
    will fuel the need for a low-cost web terminal.  StrongARM
    represents the kind of technology that will put this product
    within reach of the mass consumer market."
        
    PDAs AND ORGANIZERS
        
        The SA-110 StrongARM processor will greatly enhance the
    functionality of next generation PDAs and electronic
    organizers.  "Apple's Newton team and the StrongARM design
    team have worked closely together during the past eighteen
    months," said Michael Culbert, system architect, Apple
    Computer, Inc.  "We are very excited about this new
    technology and its potential to carry the next generation of
    Newton PDAs to a new level.  Our customers and licensees
    will be delighted by the new applications and human
    interface capabilities this chip can enable."
        In addition, application developers targeting mobile
    workers are porting key applications to the StrongARM
    platform.  Papyrus Associates Inc., a leader in handwriting
    recognition software has endorsed the StrongARM technology.
    "The excellent computational capabilities will enable us to
    offer improved handwriting recognition software," said Bill
    Kania, president, Papyrus Associates Inc.  "Thanks to a
    mature compiler environment, our software was easily ported
    to StrongARM."
        Dragon Systems, Inc., the industry leader in speech
    recognition technology, is also enthusiastic about the
    StrongARM technology.  "The performance delivered by the
    SA-110 will enable Dragon Systems to provide advanced speech
    recognition capabilities for handheld portable products,"
    said Stephen Breit, manager of special projects, Dragon
    Systems.
        
    SET-TOP BOXES AND VIDEO GAMES
        
        Second generation set-top boxes will drive the movement
    to real interactive TV and hyper-realistic 3D video games in
    the $200-$400 range. "Interactive set-top boxes are a
    demanding product to build -- you need twice the performance
    of a desktop PC at one third the cost," said Malcolm Bird,
    chief executive, Online Media, Ltd., a leading supplier of
    set-top box technology and products.  "While the performance
    of these StrongARM processors is impressive, what sold us on
    this technology is the price points at which the performance
    is delivered.  This technology will help make interactive TV
    a reality."
        
    INDUSTRY LEADING MIPS/WATT AND MIPS/DOLLAR
        
        The SA-110, available in 100 MHz, 160 MHz, and 200 MHz
    internal clock speeds, has set new industry benchmarks in
    terms of both power- and cost-efficiency, as well as overall
    processor performance.
        The 100 MHz part operating at 1.65 volts, delivers 115
    Dhrystone 2.1 MIPS while dissipating less than 300 mW of
    power.  The 160 MHz version delivers 185 Dhrystone 2.1 MIPS
    at only 450 mW of power dissipation, giving a
    performance/power ratio (MIPS-per-watt) of over 400.  This
    makes it the most power-efficient processor available today.
        The 200 MHz part operates at 2.0 volts and performs 230
    Dhrystone 2.1 MIPS while still running on under a watt of
    power.  This device, priced at under $50, achieves the
    industry's best performance/price ratio at almost 5 MIPS/$
    (MIPS-per-dollar).  Pricing for the 100 MHz part is less
    than $29 in 10k unit quantities, while the 160 MHz part is
    available for $49 in the same quantities.
        Cost reduction is a primary focus for the SA-110
    product.  All three versions are packaged in a low-cost,
    small footprint, plastic package (144-pin plastic TQFP).
    The SA-110 can accommodate 3.3 volt input/output levels,
    allowing system designers to utilize off-the-shelf 3.3 volt
    memories and other commodity components.
        The SA-110 is produced on eight-inch wafers on a 0.35
    micron CMOS process at Digital's state-of-the-art Fab 6
    facility in Hudson, Mass.  Samples are available now, with
    production scheduled to begin in the spring.  Software
    development tools (compilers, assemblers, debuggers),
    operating systems, and applications are available through
    Digital and other third party companies supporting the ARM
    architecture.
        Digital Semiconductor, a Digital Equipment Corporation
    business headquartered in Hudson, Massachusetts, designs,
    manufactures and markets industry-leading semiconductor
    products including Alpha microprocessors and PCI chips for
    networking, bridging, and graphics/multimedia, as well as
    low-power StrongARM microprocessors under license from
    Advanced RISC Machines Ltd.  Mitsubishi Electric Company is
    a second source for Alpha microprocessors.  World Wide Web
    site: http://www.digital.com/info/semiconductor
         Digital Equipment Corporation is the world's leader in
    open client/server solutions from personal computing to
    integrated worldwide information systems.  Digital's
    scalable Alpha platforms, storage, networking, software and
    services, together with industry-focused solutions from
    business partners, help organizations compete and win in
    today's global marketplace.
    

    Re: Internet Terminals.
    
    
    The current issue of PCWEEK (Feb5) describes WYSE' new Winterm Model
    2000 as a potential foray into the Internet Terminal "Market". This 
    device acts a display for CITRIX' multiuser variant of NT (processing
    is done at a single NT server which supports multiple displays). Put 
    a browser on the server and voila!, $500-$750 Internet Terminal. 
    
    And now, a moment of rambling about secret things from days gone
    by... things I seldom have opportunity to ramble about :-)...
    
                             *  *  *
    
     We built a prototype of one of these roughly 3 years ago following
    the exact same approach... refined it a bit 1.5 years ago by burning it 
    into a cartridge for a VTLAN40 about 1.5 years ago. I didn't use the 
    Internet with it, but did run an NT display at home via a 9600 baud dialup 
    to a 486 running CITRIX NT. 
    
    Dave Doucette (former co-worker at C&P) wrote a white paper about
    creating a business by providing Alpha-based CITRIX NT servers and 
    providing subscribers with VTLAN40 cartridge-based displays for access
    to applications and Internet. 
    
    The VTLAN40 platform was designed for quick product creation by 
    burning software into it's "Nintendo" cartridges, we began tinkering
    with the challenge of burning in a browser and making what we were 
    calling a "dedicated web terminal"... this was about 1.5 years ago...
    but we didn't go anywhere with it.  
    
    
    								- Mac

I'm probably covering old ground here, but is the ARM processor the thing 
that Acorn developed for their Archimedes series?

Chris.

    re .138:
    
    Citrix still doesn't have an Alpha version of their NT... :-(
    
    I have it up and running here on a Pentium machine. Clients are now
    being built (apart from Wyse) at least by NCD and Tektronix. Insignia
    has licenced the base sofwtare too.
    

The attached is from an open Internet discussion group.....
    
                      I N T E R O F F I C E   M E M O R A N D U M

                                        Date:     08-Feb-1996 06:02am GMT 
                                        From:     Bobby Seder
                                                  [email protected]@umc@VALMTS@VBO
                                        Dept:      
                                        Tel No:    

TO:  Kirk Fleming/Metro Technologies      ( [email protected]@internet )
TO:  [email protected]@internet

Subject: Re: Alpha/NT vs Intel/NT

At 23:13 02-07-96 +0000, Kirk R Fleming wrote:
>I'm a new user of this list--please let me know if I'm using it improperly.
>
>I'm interested in getting as much objective data regarding the 
>cost/performance advantages of NT-Alpha machines vs. NT-Intel machines, 
>especially SMP units in the $20-30K range, if they indeed exist.
>
>Also, if anyone is aware of any reliability comparisons between Intel and 
>Alpha boxes in the Compaq ProLiant 4500 class (if they exist), I'd greatly 
>appreciate it.

Um... Beware, you will probably get 10 zillion responses to this one, that's
like asking if Ford is Better than Chevy or Pepsi better than Coke.

We have about 25 production servers, six of which are Alpha's (4 AlphaServer
2100's and 2 AlphaServer 1000's, oh yeah, my workstation is an Alpha as well
about 16 people in our co. have AlphaStations) the rest are Compaq's. We
could go into discussion for days with this one, but here is the bottom line
for me:

Setup and recovery:
-------------------
Intel is fairly painful...
To set up you have to *hope* your CD is supported, if not, you have to load
DOS and load the CD driver... God forbid if you lose a drive, there are very
few RAID controllers that run well AND have reliable utilities.

Alpha:
EVERYTHING is SCSI. EVERYTHING either works or it dosen't - I have never run
into any SCSI device that wasn't supported on an Alpha. To install, you plop
the CD in and GO. No loading drivers or anything. There are a LOT of GOOD
RAID controllers that run on Alpha, the best (I think) is the DEC SWXCR
card. It has ARC-based utilities (i.e. you don't have to have an OS running
to fix your drives).

Stability:
----------
Intel:
Intel backplane is a 32-bit bus. It has less advanced memory parity
checking. In the past year we've had probably 3 dozen Intel crashes (most of
them memory-based error's of pagefile I/O prblms).

Alpha:
In the year that I've been here we have not had *ONE* crash on an Alpha.
Period. They're like Energizer bunnies.

Performance:
------------
Intel:
Intel is 32-bit based and is a CISC (Complex Instruction Set Chipset) based
CPU. CISC processors work (more or less) by putting a zillion often-used
instructions hard-coded on the chip. This is very efficient (because there
is no translation) so long as you ARE doing those instructions, the burden
of efficiency is on the CPU. An (very weak) analogy is what if you only need
to go to the end of your driveway, but your car only knows how to go to the
store and back - well that isn't very efficient. Event the brand-new
PentiumPro (P6) is 32-bit. Windows NT is still 32-bit but will be 64-bit
probably in v5.0. 32bit means 32bits of information are being transferred at
a time across the bus (motherboard) and the CPU.

Alpha:
The Alpha is a 64-bit based backplane and CPU. Also, the Alpha is a RISC
(Reduced Instruction Set Chipset) CPU. The RISC processors wotk with just a
few instructions (7, I think) hardcoded on the CPU. Now, the burden of
efficiency is on the compiler (which we now have - and didn't have in the
bad old days!). So RISC turns out to be faster in nearly every test
(benchmark). Secondly, since the chip isn't filled with a zillion
instructions, the CPU has PLENTY of room for level1 cache (right on the CPU
very fast) and room for faster crystals (like the 350MHz that just came
out). Although NT can only utilize 32-bit right now, It will be a long time
before NT outgrows this box. You will never have to worry about bus contention.

Future:
-------
Intel:
Intel is still making CISC processors - for the PC market. That really is
the place for Intel. From a server management side and business side, they
are costly and proven not as stable as Alpha.

Alpha:
Alpha is still a baby (5 yrs old) and has nearly incomprehensible potential.
Alpha is CLEARLY the direction of Microsoft (and DEC and US, PacifiCare)...

B O T T O M  L I N E:
---------------------
I have had nothing but great experiences with the Alpha's... They are great
to work on and painless to manage... Especially in comparision to the Intel.
Compaq's? Don't EVEN get me started.. I would't accept a Compaq, no a
TRUCKLOAD of Compaq's if you GAVE them to me - talk about a managment nightmare!


WHEW! you caught me in a talkative mood (it's 04:00). I had to do the
justification HERE for the strategic direction toward Alpha. There is a lot
of stuff on the net about processors and benchmark comparision. You can
learn a lot from what's out there (I learned a LOT about CPU's).

Hope this helps! Let me know if you have any questions. Please don't post
(or reply this to the list) as I do not want to get involved in this (what
will probably be a battle).

Thanks.

---------------------------------------------------------------------------
Bobby Seder MCPS
Windows NT Systems Manager              Internet
PacifiCare Health Systems                       http://www.phs.com
5995 Plaza Drive M/S 3101               Email:
Cypress, CA  90630                              [email protected]
714.236.7148                            PGP:
714.236.7882 fax                                finger [email protected]
     Key fingerprint =  D6 66 95 5B 67 C1 1D C0  4B 1D 43 4A 67 95 80 8D
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

    Hmmm. A RISC chip with 7 instructions, that is reduced.
    
    Still good to hear that people love us again. Thanks Chris.
    
    Andrew

    re: .143
    
    7 instructions?  Well, the PDP-8 got by with only 5 memory reference
    instructions.  Our first RISC processor?!  :-)  

Andrew:

> Hmmm. A RISC chip with 7 instructions, that is reduced.

  Well, he probably got AND, TAD, DCA, JMP, JMS, IOT, and OPERATE,
  but forgot ISZ. :-)

                                   Atlant

    Opps - you're right.  6 memory reference instructions!  :-)
    
    (Plus the IOTs and OPRs, of course.)
    

RE: .142

Someone should sign this guy up to do testimonials.  His is by far one of
the most articulated comparisions of Intel Vs. Alpha that I've seen by a
real live user.  This is good stuff.

    Well, a nice testimonial, but filled with innacuracies:
    
    1) "Alpha: everything is SCSI, Intel not."  You can always order
      an Intel with all SCSI, and if you wait a bit, Alpha will have
      IDE-related nightmares as well ;-)  Intel's support of SCSI is
      much greater, especially since now several SCSI boards are
      being pulled from the Alpha Windows NT HCL (due to needing Intel
      to run config programs).
    2) "Intel backplane is 32bit, Alpha is 64bit."  Well, some Alphas
      have a 64bit PCI, but most have 32bit.  All P6 models have a
      64 bit path to memory; the Alpha XL's are 32 bit, for example.
    3) "RISC turns out faster in nearly every test."  Hogwash.  Check
      out  http://www.bapco.com, for example.  Alpha *just* passed
      P6 with SPECint95, and take a look at the price of the Alpha
      that it took to do it.  Compare to the price of a Gateway G6 200.
      Anything Ev45-based, such as the Alpha XL, gets kicked around
      by P6, let alone faster Pentiums.  Look at the performance of
      Alpha-powered Multia vs. Intel-powered Multia.  Alpha-Multia
      gets it's ass absolutely kicked.
    4) "Alpha CPU has plenty of room for L1 cache".  Huh?  I got lost
      on this one.  But if you want to compare cache size and cost:
      P6 has L1 *and* 256k (or 512k) L2 cache BUILT IN.  No P6 machine
      has or needs external cache.  Alpha requires LOTS (at least 2Mb)
      of separate and costly L2 cache to perform.  Hardly a win for
      Alpha system designers and customers who must pay more just to
      obtain the same level of performance.
     5) P6 boxes have ECC memory.  Alpha just has single word parity
      in the AlphaStations.
    
    .02 Kratz

    Methink Mr. Kratz is picking his targets carefully.  He compares the
    Intel Multia to the Alpha Multia, based on the 21066 chip, which began
    shipping two years ago, before the Pentium 60 hit the streets, and with 
    its integral PCI and memory controller was never considered a truly 
    high-performance chip.  The EV45, based on a previous generation of the 
    Alpha architecture, has been shipping in volume for 19 months, whereas 
    the P6 has been around for barely two months.  The Alpha XL is built 
    on a PC chassis meant primarily to support Intel X86 chips, so its "32-bit 
    bus" is characteristic of all XL systems, not "Alpha."  And as for
    "just" passing the P6 in SPEC95 performance, the Alpha 21164-333,
    announced last October and shipping since December, does that handily
    for SPECint95, while for SPECfp95 there is absolutely no comparison;  the
    P6 falls flat.  And for SPEC92, Intel recently released new test results 
    for P6 with a performance *reduction* of nearly 15%.
    
    M
    
    Of course, the "testimonial" concerned servers, not Multias or XLs.  How
    does he think Alta Vista would perform with Intel servers? 

>    Methink Mr. Kratz is picking his targets carefully.

Personally, I think *we're* (Alpha in this context) the ones "picking
targets carefully".  Yes, an Alpha ev5/350 *turbolaser* and ev5/300+
expensive workstations can beat a PP200 Intel system on benchmarks.
Great.  (Insert cynical face here...)

[further diatribes on Alpha pricing supressed]

    And as for
    "just" passing the P6 in SPEC95 performance, the Alpha 21164-333,
    announced last October and shipping since December, does that handily
    for SPECint95, while for SPECfp95 there is absolutely no comparison;  the
    P6 falls flat.  And for SPEC92, Intel recently released new test results 
    for P6 with a performance *reduction* of nearly 15%.

As a compiler developer, I get "rather irritated" (understatement) when
we trumpet Specmarks for incomparably priced systems and fail to go after
mindshare opportunities that won't come again.

    ...and while EV45 may have been shipping for 19 months, it wasn't
    until Alpha XL shipped in October of 1995 (i.e. just before P6)
    that this company made any effort to come out with a "reasonably"
    priced Ev45 box.  I remember getting into trouble with Workstation
    management for posting a note on how to assemble parts for a "Poor
    Man's Avanti" in the AlphaStation notes file in October of 1994;
    it took *12 FULL MONTHS* for that to get packaged into a system and
    get out the door.  By the time the Alpha XLs made it out, the product
    had a lifespan measured in days before P6 came out.  Kratz

    Re .151:
    
    ...and I understand it took plenty of arm-torquing to make it happen
    even then.
    
    M

    ref . 144
    thats what we said when we looked at MIPS RISC -
    just like a PDP 8 - link & store - enhance with cache and paging
    (TLB)..
    
    remembering Palo Alto's PMAX days.. - - what would have happend if we
    would have pushed PDP8 chip designs to 64 bits ?? :^) -- :^(

    re .149  How does he think Alta Vista would perform with Intel servers? 
    
    See the March Computer Shopper for a review of web servers.  Alpha
    really stunk up the review, having significantly less performance
    than a Pentium 75's with half as much memory (16Mb vs. Alpha's 32).
    For about the same money, the Pentium 75's also included things like
    a 2Gb tape backup or 21 inch monitor (Alpha included neither a monitor
    nor tape backup).  Computer Shopper was surprised at the "less than
    expected performance".  They also commented that faster Alphas are
    available but at "significantly higher cost".
    
    Not exactly Alta Vista, but nevertheless Alpha didn't exactly show
    great internet server performance in this review, even when compared
    to the slowest CPU chip Intel makes.
    
    P.S. This was a 166Mhz LCA... but Alpha is Alpha is Alpha as far as
    the press goes.  Digital Semi makes no effort to differentiate Alphas
    and/or to try and lower expectations, so it's hard to feel sorry for
    LCA getting beat up in a review that consequently makes all Alphas
    look bad.
    .02 Kratz

Exactly what Alpha web server was reviewed in Computer Shopper?

Paul

    Sounds like it was the Multia (woof!) webserver package.  Man can't the
    Components folks let LCA Alpha die quiety rather than try to repackage
    it like this.

> Man can't the Components folks let LCA Alpha die quiety rather
> than try to repackage it like this.

  Well, you see, it's like this:

    1. We actually *BELIEVED* that LCA would perform as originally
       claimed. (At least, this is what I've been told. I didn't
       work for the Components Group at the time the claims were
       made.)

    2. We sold customers on the concept and they bought it.

    3. LCA didn't meet its original goals. Larry Cabrinety's
       head rolled, partly on the failure of Multia to perform.

    3. But the customers who bought the concept still expect us to
       continue to sell them the systems they've standardized on.


  Unreasonable, aren't those customers? Why can't they be like the
  people who buy just one PC, so you can sell them whatever hot box
  happens to your latest-and-greatest?

  On the other hand, new Multias are Intel-based and run Intel-
  binaries a good deal better than the old LCA-based multias.

  I assure you, the Components Group hasn't got any new designs
  based on the LCA you're discussing.

                                   Atlant

    Low-End AlphaServer News                             16th February 1996
    
    
            *** WNT Internet AlphaServer - Product of the Week ***
    
        *** Low-End AlphaServer Industry Standard Benchmark figures ***
    
    
    Server colleagues,
    
    Included in this issue of Low-End AlphaServer News is an an article, 
    reproduced without permission, from Network Week's WWW page on 
    http://www2.emap.com/nww/wans/alphaweb.html. This article featured on 
    the front page of this weeks Network Week, in the "Product of the Week" 
    feature slot. I have also included a reference document of the 
    industry-standard benchmark results for the AlphaServer 400 4/233,
    AlphaServer 1000A 4/233, 1000/1000A 4/266, 2000/2100 4/233, and
    2000/2100/2100A 4/275.
    
    Best regards,
    David Kerrell
    Low-End AlphaServer Marketing

    
Alpha's a little Web winner

Digital has, for some time, been marketing a Unix-based Alpha Web server.
Alongside this established machine is a new sibling - an equivalent offering 
based on Microsoft's Windows NT Server. 

The model Network Week looked at is the 4/233 version of the Alpha 400-series.
The box itself is a floor-standing beast, approximately a foot and a half tall, 
which sits neatly and relatively quietly under a desk. Unlike many Unix-style 
workstations, the monitor connector is a standard VGA socket (no expensive 
monitors, thanks very much) with the keyboard and mouse connectors looking 
suspiciously like their PS/2 equivalents. 

The box itself looks just like a PC - except for the mini-SCSI outlet on the 
back. Our server had an EtherWORKS Turbo PCI Ethernet card, with all of the 
usual connector types provided, and 32MB of internal RAM shipped with the 
machine (the theoretical maximum is 192MB, enough for most Web purposes). 

Additionally, the documentation hints that 'fast' SCSI operation is possible 
with disk units connected via the external SCSI port, though this wasn't tested 
as part of our review. 

        The software shipped with our unit comprised Windows NT
Server and Process Software's 'Purveyor'. Although NT was pre-installed, various
location-specific parameters (network addresses and so on) had to be set. As a
complete NT novice, I was pleasantly surprised that the things that needed 
changing were in completely obvious places, and that everything was configured 
within minutes (if only Windows 95 or Unix were that easy). 

This is one of the big advantages of the NT-based variant of Digital's Web 
server - the administrator, who would be frightened by a Unix front end, sees a 
user interface he or she understands. The only oddity we observed was that the 
EtherWORKS card was reluctant to operate with our 10Mbps UTP connections in 
'auto-detect' mode. This was quickly rectified through the Control Panel, 
however, by simply forcing the driver to work with the 10BaseT interface. 

While our test looked at the Purveyor Web server software, this is by no means 
the only package which will run on such a system. Users wanting to use 
alternatives, such as Netscape's Commerce or Communications servers, can do so 
with no problem. The NT versions of the Netscape products are significantly 
cheaper than their Unix equivalents too. 

Purveyor itself ships on nine disks or one CD-ROM (both are included in the
package). Installation is simple, and configuring the server is similarly
straightforward - as one would expect with anything under a Windows-based GUI. 

The package includes a single sheet 'idiot's guide', along with a user manual, 
security handbook and programmer's guide. 

Purveyor, like many other such packages, provides the network manager with an
Application Programming Interface (API), which can be used to provide
functionality over and above the more standard CGI (Common Gateway Interface)
features which are also available with Purveyor. 

The Programmer's Guide deals with how to work with these customisation features,
though at just 56 pages, don't expect it to tell you absolutely everything you 
need to know. Even more slimline is the 42-page Guide to Server Security, which 
deals with the operation of the access control features of the software. This 
manual details the commands required to run the server under an unprivileged 
account (potentially important for security reasons), along with the process for 
defining access control lists. The latter can be terribly long-winded on many 
Web servers (especially the Public Domain Unix ones, but even some under Windows 
or MacOS can require thebuilding of cryptic-looking text files), but the 
intuitive Windows front end makes it a friendly process. 

The access control mechanism also provides the network manager with a report of
incoming connections which were denied for some reason. The actual 
authentication of incoming connections can be done either by username and 
password, or by the IP address of the originating computer. 

The User's Guide is thicker than the other two manuals, and describes the 
operation of the server software itself - the general configuration and 
day-to-day operation - along with a few words on the HTTP (Web transport) 
protocol. 

The software includes many useful features such as a Wizard-based ODBC
connectivity application (which enables users to easily develop customised Web
forms and use them to access, for example, a Microsoft Access database via the
Web), and a link checker. The latter is incredibly useful, and is used to verify 
that the links in the documents on the server actually point to something (it's 
easy to mistype a URL if you're writing a Web document). 

Beside these features, which one might class as 'extras', Purveyor is similarly 
up to speed on features which are becoming pretty standard, but which aren't yet 
included in many server packages. These include: server-side includes; proxy 
server capability (including proxy-to-proxy operation); searching local WAIS 
databases; remote server management; multiple virtual servers on one machine. 

To be absolutely honest, it is terribly hard to get excited about this product. 
This is, however, because it just works, and works boringly well. The machine is 
fast enough, the software (both operating system and supplied Web server) is 
easy to use and has most of the features one might want, and the documentation 
is almost certainly slimline because everything is just so easy to handle. 

Anyone who has ever set up something like WebSTAR (formerly MacHTTP) on a
Mac, and who can relate to the fact that it just sits there and does the job, 
will understand how I feel about this Digital box - and anyone who has fought 
with Unix-based Web server config files would be envious of just how simple the 
Digital product is. Not only this, but software like Netscape's server offerings 
costs far less under NT than under Unix. This Alpha/NT product is a winner. 


Pro

   Easy to set up and administer 
   Does the job well 
   NT is far easier than Unix for system managers with little or no Unix
   experience 

Con

   It's really hard to think of any 

Price

   �6,200 

Contact

   Digital, Tel: 01734 202279

Others to consider

   Numerous PD and ShareWare packages, along with packages such as
   Netscape�s servers, which are available on most Unix platforms and under NT
   or BSDI on Intel boxes 


David Cartwright

Digital Equipment Corporation/ January 29, 1996
Performance Flash
Digital AlphaServer Family
Digital UNIX(r)

The industry-standard benchmark results for the AlphaServer 400 4/233,
AlphaServer 1000A 4/233, 1000/1000A 4/266, 2000/2100 4/233, and
2000/2100/2100A 4/275, are shown in the following table.

Digital's Low-End AlphaServer Family Digital UNIX Benchmark Results
-------------------------------------------------------------------
                               AlphaServer                  
                                                       2000/2100
Benchmarks and      400   1000A  1000/1000A  2000/2100    /2100A
   Metrics        4/233   4/233       4/266      4/233     4/275
-------------------------------------------------------------------
SPEC CINT95                                          
 SPECint95           nr      nr       4.34        nr         nr
 SPECint_base95      nr      nr       3.96        nr         nr

SPEC CFP95                                           
 SPECfp95            nr      nr       6.03        nr         nr
 SPECfp_base95       nr      nr       5.68        nr         nr

SPEC CINT95 Homogeneous Capacity
 SPECint_rate95      nr      nr       39.0        nr         nr
 SPECint_rate_base95 nr      nr       35.6        nr         nr

SPEC CFP95 Homogeneous Capacity
 SPECfp_rate95       nr      nr       54.3        nr         nr
 SPECfp_rate_base95  nr      nr       51.1        nr         nr

SPEC CINT92                                          
 SPECint92        161.0   165.3      197.5     177.3      202.9
 SPECbase_int92   141.8   153.4      178.5     163.7      187.8

SPEC CFP92                                           
 SPECfp92         194.3   222.9      263.8     215.0      292.6
 SPECbase_fp92    182.4   210.8      247.3     192.3      259.5

SPEC CINT92 Homogeneous Capacity
 SPECrate_int92                                                
   4 CPUs            na      na         na    15,538     18,036
   2 CPUs            na      na         na     8,284      9,423
   1 CPU          3,885   3,989      4,921     4,135      4,711
SPECrate_base_int92
   4 CPUs            na      na         na    14,494     16,963
   2 CPUs            na      na         na     7,367      8,617
   1 CPU          3,425   3,684      4,224     3,842      4,423

SPEC CFP92 Homogeneous Capacity
 SPECrate_fp92                                                 
   4 CPUs            na      na         na    17,361     25,997
   2 CPUs            na      na         na     9,676     13,242
   1 CPU          4,606   5,356      6,283     5,112      6,827
SPECrate_base_fp92
   4 CPUs            na      na         na    15,741     24,273
   2 CPUs            na      na         na     8,605     12,373
   1 CPU          4,340   4,979      5,907     4,575      6,182

SPEC SFS Release 1.0
2 CPUs                                                        
   SPECnfs_A93       na      na         na        nr      1,798
     (ops/sec)
   Average           na      na         na        nr       26.0
     Response Time (msec)



                               AlphaServer                  
                                                      2000/2100
Benchmarks and      400   1000A  1000/1000A  2000/2100   /2100A
   Metrics        4/233   4/233       4/266      4/233    4/275
-------------------------------------------------------------------
1 CPU                                                         
   SPECnfs_A93       nr      nr         nr     1,210         nr
     (ops/sec)
   Average           nr      nr         nr      16.3         nr
    Response Time (msec)

LINPACK 64-bit Double-Precision
 100x100 MFLOPS    46.8    44.1       50.1      50.1       51.5
 1000x1000 MFLOPS
   4 CPUs            na      na         na     407.1      642.7
   2 CPUs            na      na         na     249.9      347.9
   1 CPU          138.4   169.0      193.2     159.4      204.8

DN&R Labs CPU2                                                
 MVUPS            398.4      nr      455.5     350.7      367.6

Dhrystone                                                     
 instructions/second
  V1.1          434,185 410,493    474,303   442,320    456,184
  V2.1          416,666 384,615    454,546   416,667    454,545

SLALOM                                                        
 Patches          8,146   7,902      8,908     8,134      8,814
 MFLOPS            83.6    78.3       99.7      82.3       95.7

Livermore Loops
 Geo. mean         36.0    35.9       40.9      37.8       40.5
 MFLOPS

CERN                                                          
 CERN units        46.2    43.5       53.5      45.9       53.1

Whetstone                                                     
 KWIPS
Single-precision  267.5   251.1      303.1     256.2      259.4
Double-precision  249.3   220.8      281.5     225.7      240.0

AIM Suite III                                                 
 4 CPUs                                                        
  Performance        na      na         na     345.3         nr
    Rating
  Maximum User       na      na         na     2,231         nr
    Loads
  Maximum            na      na         na   3,383.8         nr
    Throughput (jobs/minute)





                               AlphaServer                  
                                                      2000/2100
Benchmarks and      400   1000A  1000/1000A  2000/2100   /2100A
   Metrics        4/233   4/233       4/266      4/233    4/275
-------------------------------------------------------------------
  2 CPUs                                                        
   Performance       na      na         na        nr     227.3�
    Rating
   Maximum User      na      na         na        nr     1,892�
    Loads
   Maximum           na      na         na        nr   2,227.1�
    Throughput (jobs/minute)

  1 CPU                                                          
    Performance      nr      nr         nr     132.4     156.5�
     Rating
    Maximum User     nr      nr         nr     1,031     1,298�
     Loads
    Maximum          nr      nr         nr   1,297.5   1,533.9�
     Throughput (jobs/minute)

AIM Suite VII-Multiuser Shared System
 4 CPUs                                                        
   Peak              na      na         na        nr    1,160.2�
    Performance Multiuser Jobs/Min
   Sustained         na      na         na        nr      1,100�
    Performance Multiuser Loads
   Job Timing        na      na         na        nr         89�
      Index

 1 CPU                                                         
   Peak           310.0      nr     452.1*        nr         nr
    Performance Multiuser Jobs/Min
   Sustained      283.6      nr     442.9*        nr         nr
    Performance Multiuser Loads
   Job Timing        76      nr       931*        nr         nr
    Index

Rhealstone                                                    
 Task Switch        8.4     7.7        6.6       7.2         nr
  (�sec)
 Preemption        19.5    16.3       16.1      14.1         nr
  (�sec)
 Message           35.0    43.1       34.7      36.2         nr
  Passing (�sec)
 Semaphore         71.0    71.8       59.1      70.9         nr
  Shuffle (�sec)precision
-------------------------------------------------------------------
na = not available; nr = not reported
� Result valid for AlphaServer 2000/2100 4/275 2 CPUs configuration only.
� Result valid for AlphaServer 2000/2100 4/275 1 CPU configuration only.
� Result valid for AlphaServer 2100 4/275 4 CPUs configuration only.
* Result valid for AlphaServer 1000 4/266 configuration only.
Note: AlphaServer 2000 systems support a maximum of 2 CPUs; AlphaServer 
2100/2100A systems support a maximum of 4 CPUs.
-------------------------------------------------------------------


Digital evaluated the performance of the AlphaServer family using 
industry-standard benchmarks. These benchmarks allow comparisons across
vendors' systems. Performance characterization is just one "data point" to
be used in conjunction with other purchase criteria such as features, 
services, and price.

For more information on Digital's AlphaServer family, contact your local 
Digital sales representative. Please send questions and comments about the 
information presented in this Performance Flash to Internet address:
[email protected].

Copyright � 1996 Digital Equipment Corporation. All rights reserved. 

Digital, the Digital logo, the AlphaGeneration logo, and AlphaServer are 
trademarks of Digital Equipment Corporation.

SPEC, SPECint92, SPECfp92, SPECrate_int92, and SPECrate_fp92 are trademarks of 
the Standard Performance Evaluation Corporation. UNIX is a registered trademark 
in the United States and other countries, exclusively licensed through
X/Open Company Ltd. AIM is a trademark of AIM Technology, Inc. TPC-C Benchmark 
and tpmC are trademarks of the Transaction Processing Performance Council.